@aws-amplify/ui-react-liveness 3.4.7 → 3.5.1

package/dist/esm/components/FaceLivenessDetector/FaceLivenessDetectorCore.mjs

@@ -3,7 +3,7 @@ import { useInterpret } from '@xstate/react';
 import { livenessMachine } from './service/machine/machine.mjs';
 import './service/types/liveness.mjs';
 import '@tensorflow/tfjs-core';
-import '@tensorflow-models/face-detection';
+import '../../node_modules/@tensorflow-models/face-detection/dist/face-detection.esm.mjs';
 import '@tensorflow/tfjs-backend-wasm';
 import '@tensorflow/tfjs-backend-cpu';
 import '@aws-amplify/core/internals/utils';

package/dist/esm/components/FaceLivenessDetector/LivenessCheck/LivenessCameraModule.mjs

@@ -5,7 +5,7 @@ import { useColorMode } from '@aws-amplify/ui-react/internal';
 import '../service/machine/machine.mjs';
 import { FaceMatchState } from '../service/types/liveness.mjs';
 import '@tensorflow/tfjs-core';
-import '@tensorflow-models/face-detection';
+import '../../../node_modules/@tensorflow-models/face-detection/dist/face-detection.esm.mjs';
 import '@tensorflow/tfjs-backend-wasm';
 import '@tensorflow/tfjs-backend-cpu';
 import '@aws-amplify/core/internals/utils';

package/dist/esm/components/FaceLivenessDetector/LivenessCheck/LivenessCheck.mjs

@@ -4,7 +4,7 @@ import '../service/machine/machine.mjs';
 import '../service/types/liveness.mjs';
 import { LivenessErrorState } from '../service/types/error.mjs';
 import '@tensorflow/tfjs-core';
-import '@tensorflow-models/face-detection';
+import '../../../node_modules/@tensorflow-models/face-detection/dist/face-detection.esm.mjs';
 import '@tensorflow/tfjs-backend-wasm';
 import '@tensorflow/tfjs-backend-cpu';
 import '@aws-amplify/core/internals/utils';

package/dist/esm/components/FaceLivenessDetector/service/machine/machine.mjs

@@ -19,6 +19,17 @@ import { TelemetryReporter } from '../utils/TelemetryReporter/TelemetryReporter.
 const CAMERA_ID_KEY = 'AmplifyLivenessCameraId';
 const DEFAULT_FACE_FIT_TIMEOUT = 7000;
 let responseStream;
+// Helper function to get selected device info
+const getSelectedDeviceInfo = (context) => {
+    const selected = context.videoAssociatedParams?.selectableDevices?.find((d) => d.deviceId === context.videoAssociatedParams?.selectedDeviceId);
+    return selected
+        ? {
+            deviceId: selected.deviceId,
+            groupId: selected.groupId,
+            label: selected.label ?? '',
+        }
+        : undefined;
+};
 const responseStreamActor = async (callback) => {
     try {
         const stream = await responseStream;
@@ -79,6 +90,9 @@ const responseStreamActor = async (callback) => {
         }
     }
 };
+function getLastSelectedCameraId() {
+    return localStorage.getItem(CAMERA_ID_KEY);
+}
 function setLastSelectedCameraId(deviceId) {
     localStorage.setItem(CAMERA_ID_KEY, deviceId);
 }
@@ -631,6 +645,9 @@ const livenessMachine = createMachine({
                 if (event.data.message.includes('15 fps')) {
                     errorState = LivenessErrorState.CAMERA_FRAMERATE_ERROR;
                 }
+                else if (event.data.message.includes(LivenessErrorState.DEFAULT_CAMERA_NOT_FOUND_ERROR)) {
+                    errorState = LivenessErrorState.DEFAULT_CAMERA_NOT_FOUND_ERROR;
+                }
                 else {
                     errorState = LivenessErrorState.CAMERA_ACCESS_ERROR;
                 }
@@ -640,15 +657,26 @@ const livenessMachine = createMachine({
                     state: errorState,
                     error: error,
                 };
-                context.componentProps.onError?.(livenessError);
+                context.componentProps.onError?.(livenessError, getSelectedDeviceInfo(context));
                 return errorState;
             },
         }),
         callMobileLandscapeWarningCallback: assign({
             errorState: () => LivenessErrorState.MOBILE_LANDSCAPE_ERROR,
         }),
+        getSelectedDeviceInfo: (context) => getSelectedDeviceInfo(context),
         callUserCancelCallback: (context) => {
-            context.componentProps.onUserCancel?.();
+            const { onUserCancel } = context.componentProps ?? {};
+            if (!onUserCancel) {
+                return;
+            }
+            try {
+                onUserCancel();
+            }
+            catch (callbackError) {
+                // eslint-disable-next-line no-console
+                console.error('Error in onUserCancel callback:', callbackError);
+            }
         },
         callUserTimeoutCallback: (context) => {
             const error = new Error(context.errorMessage ?? 'Client Timeout');
@@ -657,14 +685,14 @@ const livenessMachine = createMachine({
                 state: context.errorState,
                 error: error,
             };
-            context.componentProps.onError?.(livenessError);
+            context.componentProps.onError?.(livenessError, getSelectedDeviceInfo(context));
         },
         callErrorCallback: (context, event) => {
             const livenessError = {
                 state: context.errorState,
                 error: event.data?.error || event.data,
             };
-            context.componentProps.onError?.(livenessError);
+            context.componentProps.onError?.(livenessError, getSelectedDeviceInfo(context));
         },
         cleanUpResources: (context) => {
             const { freshnessColorEl } = context.freshnessColorAssociatedParams;
@@ -775,10 +803,42 @@ const livenessMachine = createMachine({
     services: {
         async checkVirtualCameraAndGetStream(context) {
             const { videoConstraints } = context.videoAssociatedParams;
-            // Get initial stream to enumerate devices with non-empty labels
-            const initialStream = await navigator.mediaDevices.getUserMedia({
-                video: { ...videoConstraints },
+            const { componentProps } = context;
+            let targetDeviceId;
+            let cameraNotFound = false;
+            if (componentProps?.config?.deviceId) {
+                targetDeviceId = componentProps.config.deviceId;
+            }
+            else {
+                targetDeviceId = getLastSelectedCameraId() ?? undefined;
+            }
+            const initialStream = await navigator.mediaDevices
+                .getUserMedia({
+                video: {
+                    ...videoConstraints,
+                    ...(targetDeviceId
+                        ? { deviceId: { exact: targetDeviceId } }
+                        : {}),
+                },
                 audio: false,
+            })
+                .catch((error) => {
+                if (error instanceof DOMException &&
+                    (error.name === 'NotFoundError' ||
+                        error.name === 'OverconstrainedError')) {
+                    // Mark camera as not found when a specific target device (either provided via props
+                    // or previously selected/saved as default) cannot be accessed.
+                    if (targetDeviceId && !cameraNotFound) {
+                        cameraNotFound = true;
+                    }
+                    return navigator.mediaDevices.getUserMedia({
+                        video: {
+                            ...videoConstraints,
+                        },
+                        audio: false,
+                    });
+                }
+                throw error;
             });
             const devices = await navigator.mediaDevices.enumerateDevices();
             const realVideoDevices = devices
@@ -792,7 +852,7 @@ const livenessMachine = createMachine({
                 .getTracks()
                 .filter((track) => {
                 const settings = track.getSettings();
-                return settings.frameRate >= 15;
+                return (settings.frameRate ?? 0) >= 15;
             });
             if (tracksWithMoreThan15Fps.length < 1) {
                 throw new Error('No camera found with more than 15 fps');
@@ -811,11 +871,16 @@ const livenessMachine = createMachine({
                 });
             }
             setLastSelectedCameraId(deviceId);
-            return {
+            const result = {
                 stream: realVideoDeviceStream,
                 selectedDeviceId: initialStreamDeviceId,
                 selectableDevices: realVideoDevices,
             };
+            // If a specific camera was requested but not found, trigger a specific error
+            if (cameraNotFound) {
+                throw new Error(LivenessErrorState.DEFAULT_CAMERA_NOT_FOUND_ERROR);
+            }
+            return result;
         },
         // eslint-disable-next-line @typescript-eslint/require-await
         async openLivenessStreamConnection(context) {
@@ -1068,11 +1133,22 @@ const livenessMachine = createMachine({
             livenessStreamProvider.dispatchStreamEvent({ type: 'streamStop' });
         },
         async getLiveness(context) {
-            const { onAnalysisComplete } = context.componentProps;
-            // Get liveness result
-            await onAnalysisComplete();
+            const { onAnalysisComplete } = context.componentProps ?? {};
+            if (!onAnalysisComplete) {
+                return;
+            }
+            try {
+                const deviceInfo = getSelectedDeviceInfo(context);
+                await onAnalysisComplete(deviceInfo);
+            }
+            catch (callbackError) {
+                // eslint-disable-next-line no-console
+                console.error('Error in onAnalysisComplete callback:', callbackError);
+                // Rethrow to allow the state machine to handle the error
+                throw callbackError;
+            }
         },
     },
 });
 
-export { livenessMachine };
+export { getSelectedDeviceInfo, livenessMachine };

package/dist/esm/components/FaceLivenessDetector/service/types/error.mjs

@@ -11,6 +11,7 @@ const LivenessErrorState = {
     FACE_DISTANCE_ERROR: 'FACE_DISTANCE_ERROR',
     MOBILE_LANDSCAPE_ERROR: 'MOBILE_LANDSCAPE_ERROR',
     MULTIPLE_FACES_ERROR: 'MULTIPLE_FACES_ERROR',
+    DEFAULT_CAMERA_NOT_FOUND_ERROR: 'DEFAULT_CAMERA_NOT_FOUND_ERROR',
 };
 
 export { LivenessErrorState };

package/dist/esm/components/FaceLivenessDetector/service/utils/blazefaceFaceDetection.mjs

@@ -1,5 +1,5 @@
 import { ready, setBackend } from '@tensorflow/tfjs-core';
-import { createDetector, SupportedModels } from '@tensorflow-models/face-detection';
+import { createDetector as te, SupportedModels as Q } from '../../../../node_modules/@tensorflow-models/face-detection/dist/face-detection.esm.mjs';
 import { version_wasm, setWasmPaths } from '@tensorflow/tfjs-backend-wasm';
 import '@tensorflow/tfjs-backend-cpu';
 import { jitteredExponentialRetry } from '@aws-amplify/core/internals/utils';
@@ -32,7 +32,7 @@ class BlazeFaceFaceDetection extends FaceDetection {
         }
         try {
             await ready();
-            this._model = await createDetector(SupportedModels.MediaPipeFaceDetector, {
+            this._model = await te(Q.MediaPipeFaceDetector, {
                 runtime: 'tfjs',
                 detectorModelUrl: this.faceModelUrl,
             });

package/dist/esm/components/FaceLivenessDetector/shared/DefaultStartScreenComponents.mjs

@@ -5,7 +5,7 @@ import { CancelButton } from './CancelButton.mjs';
 import '../service/machine/machine.mjs';
 import '../service/types/liveness.mjs';
 import '@tensorflow/tfjs-core';
-import '@tensorflow-models/face-detection';
+import '../../../node_modules/@tensorflow-models/face-detection/dist/face-detection.esm.mjs';
 import '@tensorflow/tfjs-backend-wasm';
 import '@tensorflow/tfjs-backend-cpu';
 import '@aws-amplify/core/internals/utils';

package/dist/esm/components/FaceLivenessDetector/shared/FaceLivenessErrorModal.mjs

@@ -5,7 +5,7 @@ import '../service/machine/machine.mjs';
 import '../service/types/liveness.mjs';
 import { LivenessErrorState } from '../service/types/error.mjs';
 import '@tensorflow/tfjs-core';
-import '@tensorflow-models/face-detection';
+import '../../../node_modules/@tensorflow-models/face-detection/dist/face-detection.esm.mjs';
 import '@tensorflow/tfjs-backend-wasm';
 import '@tensorflow/tfjs-backend-cpu';
 import '@aws-amplify/core/internals/utils';

package/dist/esm/components/FaceLivenessDetector/shared/Hint.mjs

@@ -3,7 +3,7 @@ import { VisuallyHidden, View } from '@aws-amplify/ui-react';
 import '../service/machine/machine.mjs';
 import { FaceMatchState, IlluminationState } from '../service/types/liveness.mjs';
 import '@tensorflow/tfjs-core';
-import '@tensorflow-models/face-detection';
+import '../../../node_modules/@tensorflow-models/face-detection/dist/face-detection.esm.mjs';
 import '@tensorflow/tfjs-backend-wasm';
 import '@tensorflow/tfjs-backend-cpu';
 import '@aws-amplify/core/internals/utils';

package/dist/esm/index.mjs

@@ -3,7 +3,7 @@ export { default as FaceLivenessDetectorCore } from './components/FaceLivenessDe
 import './components/FaceLivenessDetector/service/machine/machine.mjs';
 import './components/FaceLivenessDetector/service/types/liveness.mjs';
 import '@tensorflow/tfjs-core';
-import '@tensorflow-models/face-detection';
+import './node_modules/@tensorflow-models/face-detection/dist/face-detection.esm.mjs';
 import '@tensorflow/tfjs-backend-wasm';
 import '@tensorflow/tfjs-backend-cpu';
 import '@aws-amplify/core/internals/utils';

package/dist/esm/node_modules/@tensorflow-models/face-detection/dist/face-detection.esm.mjs

@@ -0,0 +1,23 @@
+import FaceDetection from '@mediapipe/face_detection';
+import { dispose, tidy, cast, squeeze, image, expandDims, browser, Tensor, tensor1d, tensor2d, slice, add, mul, div, exp, sub, concat, reshape, sigmoid, clipByValue, util } from '@tensorflow/tfjs-core';
+import { loadGraphModel } from '@tensorflow/tfjs-converter';
+
+/**
+    * @license
+    * Copyright 2024 Google LLC. All Rights Reserved.
+    * Licensed under the Apache License, Version 2.0 (the "License");
+    * you may not use this file except in compliance with the License.
+    * You may obtain a copy of the License at
+    *
+    * http://www.apache.org/licenses/LICENSE-2.0
+    *
+    * Unless required by applicable law or agreed to in writing, software
+    * distributed under the License is distributed on an "AS IS" BASIS,
+    * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+    * See the License for the specific language governing permissions and
+    * limitations under the License.
+    * =============================================================================
+    */
+var b=function(){return b=Object.assign||function(e){for(var t,n=1,i=arguments.length;n<i;n++)for(var o in t=arguments[n])Object.prototype.hasOwnProperty.call(t,o)&&(e[o]=t[o]);return e},b.apply(this,arguments)};function T(e,t,n,i){return new(n||(n=Promise))((function(o,r){function a(e){try{h(i.next(e));}catch(e){r(e);}}function s(e){try{h(i.throw(e));}catch(e){r(e);}}function h(e){var t;e.done?o(e.value):(t=e.value,t instanceof n?t:new n((function(e){e(t);}))).then(a,s);}h((i=i.apply(e,[])).next());}))}function C(e,t){var n,i,o,r,a={label:0,sent:function(){if(1&o[0])throw o[1];return o[1]},trys:[],ops:[]};return r={next:s(0),throw:s(1),return:s(2)},"function"==typeof Symbol&&(r[Symbol.iterator]=function(){return this}),r;function s(r){return function(s){return function(r){if(n)throw new TypeError("Generator is already executing.");for(;a;)try{if(n=1,i&&(o=2&r[0]?i.return:r[0]?i.throw||((o=i.return)&&o.call(i),0):i.next)&&!(o=o.call(i,r[1])).done)return o;switch(i=0,o&&(r=[2&r[0],o.value]),r[0]){case 0:case 1:o=r;break;case 4:return a.label++,{value:r[1],done:!1};case 5:a.label++,i=r[1],r=[0];continue;case 7:r=a.ops.pop(),a.trys.pop();continue;default:if(!(o=a.trys,(o=o.length>0&&o[o.length-1])||6!==r[0]&&2!==r[0])){a=0;continue}if(3===r[0]&&(!o||r[1]>o[0]&&r[1]<o[3])){a.label=r[1];break}if(6===r[0]&&a.label<o[1]){a.label=o[1],o=r;break}if(o&&a.label<o[2]){a.label=o[2],a.ops.push(r);break}o[2]&&a.ops.pop(),a.trys.pop();continue}r=t.call(e,a);}catch(e){r=[6,e],i=0;}finally{n=o=0;}if(5&r[0])throw r[1];return {value:r[0]?r[1]:void 0,done:true}}([r,s])}}}var O=["rightEye","leftEye","noseTip","mouthCenter","rightEarTragion","leftEarTragion"];var B={modelType:"short",runtime:"mediapipe",maxFaces:1};var z=function(){function i(t){var n=this;this.width=0,this.height=0,this.selfieMode=false,this.faceDetectorSolution=new FaceDetection({locateFile:function(e,n){if(t.solutionPath){var i=t.solutionPath.replace(/\/+$/,"");return "".concat(i,"/").concat(e)}return "".concat(n,"/").concat(e)}}),this.faceDetectorSolution.setOptions({selfieMode:this.selfieMode,model:t.modelType}),this.faceDetectorSolution.onResults((function(e){if(n.height=e.image.height,n.width=e.image.width,n.faces=[],null!==e.detections)for(var t=0,i=e.detections;t<i.length;t++){var o=i[t];n.faces.push(n.normalizedToAbsolute(o.landmarks,(r=o.boundingBox,a=void 0,s=void 0,h=void 0,a=r.xCenter-r.width/2,s=a+r.width,h=r.yCenter-r.height/2,{xMin:a,xMax:s,yMin:h,yMax:h+r.height,width:r.width,height:r.height})));}var r,a,s,h;}));}return i.prototype.normalizedToAbsolute=function(e,t){var n=this;return {keypoints:e.map((function(e,t){return {x:e.x*n.width,y:e.y*n.height,name:O[t]}})),box:{xMin:t.xMin*this.width,yMin:t.yMin*this.height,xMax:t.xMax*this.width,yMax:t.yMax*this.height,width:t.width*this.width,height:t.height*this.height}}},i.prototype.estimateFaces=function(e,i){return T(this,void 0,void 0,(function(){var o,r;return C(this,(function(a){switch(a.label){case 0:return i&&i.flipHorizontal&&i.flipHorizontal!==this.selfieMode&&(this.selfieMode=i.flipHorizontal,this.faceDetectorSolution.setOptions({selfieMode:this.selfieMode})),e instanceof Tensor?(r=ImageData.bind,[4,browser.toPixels(e)]):[3,2];case 1:return o=new(r.apply(ImageData,[void 0,a.sent(),e.shape[1],e.shape[0]])),[3,3];case 2:o=e,a.label=3;case 3:return e=o,[4,this.faceDetectorSolution.send({image:e})];case 4:return a.sent(),[2,this.faces]}}))}))},i.prototype.dispose=function(){this.faceDetectorSolution.close();},i.prototype.reset=function(){this.faceDetectorSolution.reset(),this.width=0,this.height=0,this.faces=null,this.selfieMode=false;},i.prototype.initialize=function(){return this.faceDetectorSolution.initialize()},i}();function D(e){return T(this,void 0,void 0,(function(){var t,n;return C(this,(function(i){switch(i.label){case 0:return t=function(e){if(null==e)return b({},B);var t=b({},e);return t.runtime="mediapipe",null==t.modelType&&(t.modelType=B.modelType),null==t.maxFaces&&(t.maxFaces=B.maxFaces),t}(e),[4,(n=new z(t)).initialize()];case 1:return i.sent(),[2,n]}}))}))}function A(e,t,n,i){var o=e.width,r=e.height,a=1,s=Math.cos(e.rotation),h=Math.sin(e.rotation),u=e.xCenter,c=e.yCenter,l=1/t,f=1/n,d=new Array(16);return d[0]=o*s*a*l,d[1]=-r*h*l,d[2]=0,d[3]=(-0.5*o*s*a+.5*r*h+u)*l,d[4]=o*h*a*f,d[5]=r*s*f,d[6]=0,d[7]=(-0.5*r*s-.5*o*h*a+c)*f,d[8]=0,d[9]=0,d[10]=o*l,d[11]=0,d[12]=0,d[13]=0,d[14]=0,d[15]=1,function(e){if(16!==e.length)throw new Error("Array length must be 16 but got ".concat(e.length));return [[e[0],e[1],e[2],e[3]],[e[4],e[5],e[6],e[7]],[e[8],e[9],e[10],e[11]],[e[12],e[13],e[14],e[15]]]}(d)}function F(e){return e instanceof Tensor?{height:e.shape[0],width:e.shape[1]}:{height:e.height,width:e.width}}function E(e){return e instanceof Tensor?e:browser.fromPixels(e)}function R(e,t){util.assert(0!==e.width,(function(){return "".concat(t," width cannot be 0.")})),util.assert(0!==e.height,(function(){return "".concat(t," height cannot be 0.")}));}function L(e,t){var n=function(e,t,n,i){var o=t-e,r=i-n;var a=r/o;return {scale:a,offset:n-e*a}}(0,255,t[0],t[1]);return tidy((function(){return add(mul(e,n.scale),n.offset)}))}function K(e,t,n){var i=t.outputTensorSize,r=t.keepAspectRatio,a=t.borderMode,l=t.outputTensorFloatRange,f=F(e),d=function(e,t){return {xCenter:.5*e.width,yCenter:.5*e.height,width:e.width,height:e.height,rotation:0}}(f),p=function(e,t,n){if(void 0===n&&(n=false),!n)return {top:0,left:0,right:0,bottom:0};var i=t.height,o=t.width;R(t,"targetSize"),R(e,"roi");var r,a,s=i/o,h=e.height/e.width,u=0,c=0;return s>h?(r=e.width,a=e.width*s,c=(1-h/s)/2):(r=e.height/s,a=e.height,u=(1-s/h)/2),e.width=r,e.height=a,{top:c,left:u,right:u,bottom:c}}(d,i,r),m=A(d,f.width,f.height),x=tidy((function(){var t=E(e),n=tensor2d(function(e,t,n){return R(n,"inputResolution"),[1/n.width*e[0][0]*t.width,1/n.height*e[0][1]*t.width,e[0][3]*t.width,1/n.width*e[1][0]*t.height,1/n.height*e[1][1]*t.height,e[1][3]*t.height,0,0]}(m,f,i),[1,8]),o="zero"===a?"constant":"nearest",r=image.transform(expandDims(cast(t,"float32")),n,"bilinear",o,0,[i.height,i.width]);return null!=l?L(r,l):r}));return {imageTensor:x,padding:p,transformationMatrix:m}}function k(e){null==e.reduceBoxesInLowestLayer&&(e.reduceBoxesInLowestLayer=false),null==e.interpolatedScaleAspectRatio&&(e.interpolatedScaleAspectRatio=1),null==e.fixedAnchorSize&&(e.fixedAnchorSize=false);for(var t=[],n=0;n<e.numLayers;){for(var i=[],o=[],r=[],a=[],s=n;s<e.strides.length&&e.strides[s]===e.strides[n];){var h=P(e.minScale,e.maxScale,s,e.strides.length);if(0===s&&e.reduceBoxesInLowestLayer)r.push(1),r.push(2),r.push(.5),a.push(.1),a.push(h),a.push(h);else {for(var u=0;u<e.aspectRatios.length;++u)r.push(e.aspectRatios[u]),a.push(h);if(e.interpolatedScaleAspectRatio>0){var c=s===e.strides.length-1?1:P(e.minScale,e.maxScale,s+1,e.strides.length);a.push(Math.sqrt(h*c)),r.push(e.interpolatedScaleAspectRatio);}}s++;}for(var l=0;l<r.length;++l){var f=Math.sqrt(r[l]);i.push(a[l]/f),o.push(a[l]*f);}var d=0,p=0;if(e.featureMapHeight.length>0)d=e.featureMapHeight[n],p=e.featureMapWidth[n];else {var m=e.strides[n];d=Math.ceil(e.inputSizeHeight/m),p=Math.ceil(e.inputSizeWidth/m);}for(var x=0;x<d;++x)for(var g=0;g<p;++g)for(var y=0;y<i.length;++y){var v={xCenter:(g+e.anchorOffsetX)/p,yCenter:(x+e.anchorOffsetY)/d,width:0,height:0};e.fixedAnchorSize?(v.width=1,v.height=1):(v.width=o[y],v.height=i[y]),t.push(v);}n=s;}return t}function P(e,t,n,i){return 1===i?.5*(e+t):e+(t-e)*n/(i-1)}function V(e,t){var n=t[0],i=t[1];return [n*e[0]+i*e[1]+e[3],n*e[4]+i*e[5]+e[7]]}function H(e){return tidy((function(){var t=function(e){return tidy((function(){return [slice(e,[0,0,0],[1,-1,1]),slice(e,[0,0,1],[1,-1,-1])]}))}(e),n=t[0],i=t[1];return {boxes:squeeze(i),logits:squeeze(n)}}))}function U(e,t,n,i){return T(this,void 0,void 0,(function(){var i,o,r,a,u;return C(this,(function(c){switch(c.label){case 0:return e.sort((function(e,t){return Math.max.apply(Math,t.score)-Math.max.apply(Math,e.score)})),i=tensor2d(e.map((function(e){return [e.locationData.relativeBoundingBox.yMin,e.locationData.relativeBoundingBox.xMin,e.locationData.relativeBoundingBox.yMax,e.locationData.relativeBoundingBox.xMax]}))),o=tensor1d(e.map((function(e){return e.score[0]}))),[4,image.nonMaxSuppressionAsync(i,o,t,n)];case 1:return [4,(r=c.sent()).array()];case 2:return a=c.sent(),u=e.filter((function(e,t){return a.indexOf(t)>-1})),dispose([i,o,r]),[2,u]}}))}))}function j(e,t,n){return T(this,void 0,void 0,(function(){var i,s,h,u,c;return C(this,(function(p){switch(p.label){case 0:return i=e[0],s=e[1],h=function(e,t,n){return tidy((function(){var i,o,s,h;n.reverseOutputOrder?(o=squeeze(slice(e,[0,n.boxCoordOffset+0],[-1,1])),i=squeeze(slice(e,[0,n.boxCoordOffset+1],[-1,1])),h=squeeze(slice(e,[0,n.boxCoordOffset+2],[-1,1])),s=squeeze(slice(e,[0,n.boxCoordOffset+3],[-1,1]))):(i=squeeze(slice(e,[0,n.boxCoordOffset+0],[-1,1])),o=squeeze(slice(e,[0,n.boxCoordOffset+1],[-1,1])),s=squeeze(slice(e,[0,n.boxCoordOffset+2],[-1,1])),h=squeeze(slice(e,[0,n.boxCoordOffset+3],[-1,1]))),o=add(mul(div(o,n.xScale),t.w),t.x),i=add(mul(div(i,n.yScale),t.h),t.y),n.applyExponentialOnBoxSize?(s=mul(exp(div(s,n.hScale)),t.h),h=mul(exp(div(h,n.wScale)),t.w)):(s=mul(div(s,n.hScale),t.h),h=mul(div(h,n.wScale),t.h));var u=sub(i,div(s,2)),c=sub(o,div(h,2)),d=add(i,div(s,2)),p=add(o,div(h,2)),w=concat([reshape(u,[n.numBoxes,1]),reshape(c,[n.numBoxes,1]),reshape(d,[n.numBoxes,1]),reshape(p,[n.numBoxes,1])],1);if(n.numKeypoints)for(var M=0;M<n.numKeypoints;++M){var S=n.keypointCoordOffset+M*n.numValuesPerKeypoint,b=void 0,T=void 0;n.reverseOutputOrder?(b=squeeze(slice(e,[0,S],[-1,1])),T=squeeze(slice(e,[0,S+1],[-1,1]))):(T=squeeze(slice(e,[0,S],[-1,1])),b=squeeze(slice(e,[0,S+1],[-1,1])));var C=add(mul(div(b,n.xScale),t.w),t.x),O=add(mul(div(T,n.yScale),t.h),t.y);w=concat([w,reshape(C,[n.numBoxes,1]),reshape(O,[n.numBoxes,1])],1);}return w}))}(s,t,n),u=tidy((function(){var e=i;return n.sigmoidScore?(null!=n.scoreClippingThresh&&(e=clipByValue(i,-n.scoreClippingThresh,n.scoreClippingThresh)),e=sigmoid(e)):e})),[4,I(h,u,n)];case 1:return c=p.sent(),dispose([h,u]),[2,c]}}))}))}function I(e,t,n){return T(this,void 0,void 0,(function(){var i,o,r,a,s,h,u,c,l,f,d,p;return C(this,(function(m){switch(m.label){case 0:return i=[],[4,e.data()];case 1:return o=m.sent(),[4,t.data()];case 2:for(r=m.sent(),a=0;a<n.numBoxes;++a)if(!(null!=n.minScoreThresh&&r[a]<n.minScoreThresh||(s=a*n.numCoords,h=_(o[s+0],o[s+1],o[s+2],o[s+3],r[a],n.flipVertically,a),(u=h.locationData.relativeBoundingBox).width<0||u.height<0))){if(n.numKeypoints>0)for((c=h.locationData).relativeKeypoints=[],l=n.numKeypoints*n.numValuesPerKeypoint,f=0;f<l;f+=n.numValuesPerKeypoint)d=s+n.keypointCoordOffset+f,p={x:o[d+0],y:n.flipVertically?1-o[d+1]:o[d+1]},c.relativeKeypoints.push(p);i.push(h);}return [2,i]}}))}))}function _(e,t,n,i,o,r,a){return {score:[o],ind:a,locationData:{relativeBoundingBox:{xMin:t,yMin:r?1-n:e,xMax:i,yMax:r?1-e:n,width:i-t,height:n-e}}}}var N={reduceBoxesInLowestLayer:false,interpolatedScaleAspectRatio:1,featureMapHeight:[],featureMapWidth:[],numLayers:4,minScale:.1484375,maxScale:.75,inputSizeHeight:128,inputSizeWidth:128,anchorOffsetX:.5,anchorOffsetY:.5,strides:[8,16,16,16],aspectRatios:[1],fixedAnchorSize:true},W={reduceBoxesInLowestLayer:false,interpolatedScaleAspectRatio:0,featureMapHeight:[],featureMapWidth:[],numLayers:1,minScale:.1484375,maxScale:.75,inputSizeHeight:192,inputSizeWidth:192,anchorOffsetX:.5,anchorOffsetY:.5,strides:[4],aspectRatios:[1],fixedAnchorSize:true},X={runtime:"tfjs",modelType:"short",maxFaces:1,detectorModelUrl:"https://tfhub.dev/mediapipe/tfjs-model/face_detection/short/1"},Y={applyExponentialOnBoxSize:false,flipVertically:false,ignoreClasses:[],numClasses:1,numBoxes:896,numCoords:16,boxCoordOffset:0,keypointCoordOffset:4,numKeypoints:6,numValuesPerKeypoint:2,sigmoidScore:true,scoreClippingThresh:100,reverseOutputOrder:true,xScale:128,yScale:128,hScale:128,wScale:128,minScoreThresh:.5},q={applyExponentialOnBoxSize:false,flipVertically:false,ignoreClasses:[],numClasses:1,numBoxes:2304,numCoords:16,boxCoordOffset:0,keypointCoordOffset:4,numKeypoints:6,numValuesPerKeypoint:2,sigmoidScore:true,scoreClippingThresh:100,reverseOutputOrder:true,xScale:192,yScale:192,hScale:192,wScale:192,minScoreThresh:.6},G=.3,$={outputTensorSize:{width:128,height:128},keepAspectRatio:true,outputTensorFloatRange:[-1,1],borderMode:"zero"},J={outputTensorSize:{width:192,height:192},keepAspectRatio:true,outputTensorFloatRange:[-1,1],borderMode:"zero"};var Q,Z=function(){function e(e,t,n){this.detectorModel=t,this.maxFaces=n,"full"===e?(this.imageToTensorConfig=J,this.tensorsToDetectionConfig=q,this.anchors=k(W)):(this.imageToTensorConfig=$,this.tensorsToDetectionConfig=Y,this.anchors=k(N));var i=tensor1d(this.anchors.map((function(e){return e.width}))),o=tensor1d(this.anchors.map((function(e){return e.height}))),r=tensor1d(this.anchors.map((function(e){return e.xCenter}))),a=tensor1d(this.anchors.map((function(e){return e.yCenter})));this.anchorTensor={x:r,y:a,w:i,h:o};}return e.prototype.dispose=function(){this.detectorModel.dispose(),dispose([this.anchorTensor.x,this.anchorTensor.y,this.anchorTensor.w,this.anchorTensor.h]);},e.prototype.reset=function(){},e.prototype.detectFaces=function(e,t){return void 0===t&&(t=false),T(this,void 0,void 0,(function(){var n,i,r,a,s,l,p,m,x,g,y;return C(this,(function(v){switch(v.label){case 0:return null==e?(this.reset(),[2,[]]):(n=tidy((function(){var n=cast(E(e),"float32");if(t){n=squeeze(image.flipLeftRight(expandDims(n,0)),[0]);}return n})),i=K(n,this.imageToTensorConfig),r=i.imageTensor,a=i.transformationMatrix,s=this.detectorModel.execute(r,"Identity:0"),l=H(s),p=l.boxes,[4,j([m=l.logits,p],this.anchorTensor,this.tensorsToDetectionConfig)]);case 1:return 0===(x=v.sent()).length?(dispose([n,r,s,m,p]),[2,x]):[4,U(x,this.maxFaces,G)];case 2:return g=v.sent(),y=function(e,t){ void 0===e&&(e=[]);var n,i=(n=t,[].concat.apply([],n));return e.forEach((function(e){var t=e.locationData;t.relativeKeypoints.forEach((function(e){var t=V(i,[e.x,e.y]),n=t[0],o=t[1];e.x=n,e.y=o;}));var n=t.relativeBoundingBox,o=Number.MAX_VALUE,r=Number.MAX_VALUE,a=Number.MIN_VALUE,s=Number.MIN_VALUE;[[n.xMin,n.yMin],[n.xMin+n.width,n.yMin],[n.xMin+n.width,n.yMin+n.height],[n.xMin,n.yMin+n.height]].forEach((function(e){var t=V(i,e),n=t[0],h=t[1];o=Math.min(o,n),a=Math.max(a,n),r=Math.min(r,h),s=Math.max(s,h);})),t.relativeBoundingBox={xMin:o,xMax:a,yMin:r,yMax:s,width:a-o,height:s-r};})),e}(g,a),dispose([n,r,s,m,p]),[2,y]}}))}))},e.prototype.estimateFaces=function(e,t){return T(this,void 0,void 0,(function(){var n,i;return C(this,(function(o){return n=F(e),i=!!t&&t.flipHorizontal,[2,this.detectFaces(e,i).then((function(e){return e.map((function(e){for(var t=e.locationData.relativeKeypoints.map((function(e,t){return b(b({},e),{x:e.x*n.width,y:e.y*n.height,name:O[t]})})),i=e.locationData.relativeBoundingBox,o=0,r=["width","xMax","xMin"];o<r.length;o++){i[r[o]]*=n.width;}for(var a=0,s=["height","yMax","yMin"];a<s.length;a++){i[s[a]]*=n.height;}return {keypoints:t,box:i}}))}))]}))}))},e}();function ee(e){return T(this,void 0,void 0,(function(){var t,n,i;return C(this,(function(o){switch(o.label){case 0:return t=function(e){if(null==e)return b({},X);var t=b({},e);null==t.modelType&&(t.modelType=X.modelType),null==t.maxFaces&&(t.maxFaces=X.maxFaces),null==t.detectorModelUrl&&("full"===t.modelType?t.detectorModelUrl="https://tfhub.dev/mediapipe/tfjs-model/face_detection/full/1":t.detectorModelUrl="https://tfhub.dev/mediapipe/tfjs-model/face_detection/short/1");return t}(e),n="string"==typeof t.detectorModelUrl&&t.detectorModelUrl.indexOf("https://tfhub.dev")>-1,[4,loadGraphModel(t.detectorModelUrl,{fromTFHub:n})];case 1:return i=o.sent(),[2,new Z(t.modelType,i,t.maxFaces)]}}))}))}function te(e,t){return T(this,void 0,void 0,(function(){var n,i;return C(this,(function(o){if(e===Q.MediaPipeFaceDetector){if(i=void 0,null!=(n=t)){if("tfjs"===n.runtime)return [2,ee(n)];if("mediapipe"===n.runtime)return [2,D(n)];i=n.runtime;}throw new Error("Expect modelConfig.runtime to be either 'tfjs' "+"or 'mediapipe', but got ".concat(i))}throw new Error("".concat(e," is not a supported model name."))}))}))}!function(e){e.MediaPipeFaceDetector="MediaPipeFaceDetector";}(Q||(Q={}));
+
+export { z as MediaPipeFaceDetectorMediaPipe, Z as MediaPipeFaceDetectorTfjs, Q as SupportedModels, te as createDetector };

package/dist/esm/version.mjs

@@ -1,3 +1,3 @@
-const VERSION = '3.4.7';
+const VERSION = '3.5.1';
 
 export { VERSION };

package/dist/index.js

@@ -8,7 +8,8 @@ var react = require('@xstate/react');
 var uuid = require('uuid');
 var xstate = require('xstate');
 var tfjsCore = require('@tensorflow/tfjs-core');
-var faceDetection = require('@tensorflow-models/face-detection');
+var FaceDetection$1 = require('@mediapipe/face_detection');
+var tfjsConverter = require('@tensorflow/tfjs-converter');
 var tfjsBackendWasm = require('@tensorflow/tfjs-backend-wasm');
 require('@tensorflow/tfjs-backend-cpu');
 var utils = require('@aws-amplify/core/internals/utils');
@@ -23,6 +24,8 @@ var signatureV4 = require('@smithy/signature-v4');
 var uiReact = require('@aws-amplify/ui-react');
 var internal = require('@aws-amplify/ui-react/internal');
 
+function _interopDefault (e) { return e && e.__esModule ? e : { default: e }; }
+
 function _interopNamespace(e) {
     if (e && e.__esModule) return e;
     var n = Object.create(null);
@@ -42,6 +45,7 @@ function _interopNamespace(e) {
 }
 
 var React__namespace = /*#__PURE__*/_interopNamespace(React);
+var FaceDetection__default = /*#__PURE__*/_interopDefault(FaceDetection$1);
 
 /**
  * The abstract class representing FaceDetection
@@ -92,6 +96,7 @@ const LivenessErrorState = {
     FACE_DISTANCE_ERROR: 'FACE_DISTANCE_ERROR',
     MOBILE_LANDSCAPE_ERROR: 'MOBILE_LANDSCAPE_ERROR',
     MULTIPLE_FACES_ERROR: 'MULTIPLE_FACES_ERROR',
+    DEFAULT_CAMERA_NOT_FOUND_ERROR: 'DEFAULT_CAMERA_NOT_FOUND_ERROR',
 };
 
 // Face distance is calculated as pupilDistance / ovalWidth.
@@ -542,6 +547,24 @@ async function isFaceDistanceBelowThreshold({ parsedSessionInformation, faceDete
     return { isDistanceBelowThreshold, error };
 }
 
+/**
+    * @license
+    * Copyright 2024 Google LLC. All Rights Reserved.
+    * Licensed under the Apache License, Version 2.0 (the "License");
+    * you may not use this file except in compliance with the License.
+    * You may obtain a copy of the License at
+    *
+    * http://www.apache.org/licenses/LICENSE-2.0
+    *
+    * Unless required by applicable law or agreed to in writing, software
+    * distributed under the License is distributed on an "AS IS" BASIS,
+    * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+    * See the License for the specific language governing permissions and
+    * limitations under the License.
+    * =============================================================================
+    */
+var b=function(){return b=Object.assign||function(e){for(var t,n=1,i=arguments.length;n<i;n++)for(var o in t=arguments[n])Object.prototype.hasOwnProperty.call(t,o)&&(e[o]=t[o]);return e},b.apply(this,arguments)};function T(e,t,n,i){return new(n||(n=Promise))((function(o,r){function a(e){try{h(i.next(e));}catch(e){r(e);}}function s(e){try{h(i.throw(e));}catch(e){r(e);}}function h(e){var t;e.done?o(e.value):(t=e.value,t instanceof n?t:new n((function(e){e(t);}))).then(a,s);}h((i=i.apply(e,[])).next());}))}function C(e,t){var n,i,o,r,a={label:0,sent:function(){if(1&o[0])throw o[1];return o[1]},trys:[],ops:[]};return r={next:s(0),throw:s(1),return:s(2)},"function"==typeof Symbol&&(r[Symbol.iterator]=function(){return this}),r;function s(r){return function(s){return function(r){if(n)throw new TypeError("Generator is already executing.");for(;a;)try{if(n=1,i&&(o=2&r[0]?i.return:r[0]?i.throw||((o=i.return)&&o.call(i),0):i.next)&&!(o=o.call(i,r[1])).done)return o;switch(i=0,o&&(r=[2&r[0],o.value]),r[0]){case 0:case 1:o=r;break;case 4:return a.label++,{value:r[1],done:!1};case 5:a.label++,i=r[1],r=[0];continue;case 7:r=a.ops.pop(),a.trys.pop();continue;default:if(!(o=a.trys,(o=o.length>0&&o[o.length-1])||6!==r[0]&&2!==r[0])){a=0;continue}if(3===r[0]&&(!o||r[1]>o[0]&&r[1]<o[3])){a.label=r[1];break}if(6===r[0]&&a.label<o[1]){a.label=o[1],o=r;break}if(o&&a.label<o[2]){a.label=o[2],a.ops.push(r);break}o[2]&&a.ops.pop(),a.trys.pop();continue}r=t.call(e,a);}catch(e){r=[6,e],i=0;}finally{n=o=0;}if(5&r[0])throw r[1];return {value:r[0]?r[1]:void 0,done:true}}([r,s])}}}var O=["rightEye","leftEye","noseTip","mouthCenter","rightEarTragion","leftEarTragion"];var B={modelType:"short",runtime:"mediapipe",maxFaces:1};var z=function(){function i(t){var n=this;this.width=0,this.height=0,this.selfieMode=false,this.faceDetectorSolution=new FaceDetection__default["default"]({locateFile:function(e,n){if(t.solutionPath){var i=t.solutionPath.replace(/\/+$/,"");return "".concat(i,"/").concat(e)}return "".concat(n,"/").concat(e)}}),this.faceDetectorSolution.setOptions({selfieMode:this.selfieMode,model:t.modelType}),this.faceDetectorSolution.onResults((function(e){if(n.height=e.image.height,n.width=e.image.width,n.faces=[],null!==e.detections)for(var t=0,i=e.detections;t<i.length;t++){var o=i[t];n.faces.push(n.normalizedToAbsolute(o.landmarks,(r=o.boundingBox,a=void 0,s=void 0,h=void 0,a=r.xCenter-r.width/2,s=a+r.width,h=r.yCenter-r.height/2,{xMin:a,xMax:s,yMin:h,yMax:h+r.height,width:r.width,height:r.height})));}var r,a,s,h;}));}return i.prototype.normalizedToAbsolute=function(e,t){var n=this;return {keypoints:e.map((function(e,t){return {x:e.x*n.width,y:e.y*n.height,name:O[t]}})),box:{xMin:t.xMin*this.width,yMin:t.yMin*this.height,xMax:t.xMax*this.width,yMax:t.yMax*this.height,width:t.width*this.width,height:t.height*this.height}}},i.prototype.estimateFaces=function(e,i){return T(this,void 0,void 0,(function(){var o,r;return C(this,(function(a){switch(a.label){case 0:return i&&i.flipHorizontal&&i.flipHorizontal!==this.selfieMode&&(this.selfieMode=i.flipHorizontal,this.faceDetectorSolution.setOptions({selfieMode:this.selfieMode})),e instanceof tfjsCore.Tensor?(r=ImageData.bind,[4,tfjsCore.browser.toPixels(e)]):[3,2];case 1:return o=new(r.apply(ImageData,[void 0,a.sent(),e.shape[1],e.shape[0]])),[3,3];case 2:o=e,a.label=3;case 3:return e=o,[4,this.faceDetectorSolution.send({image:e})];case 4:return a.sent(),[2,this.faces]}}))}))},i.prototype.dispose=function(){this.faceDetectorSolution.close();},i.prototype.reset=function(){this.faceDetectorSolution.reset(),this.width=0,this.height=0,this.faces=null,this.selfieMode=false;},i.prototype.initialize=function(){return this.faceDetectorSolution.initialize()},i}();function D(e){return T(this,void 0,void 0,(function(){var t,n;return C(this,(function(i){switch(i.label){case 0:return t=function(e){if(null==e)return b({},B);var t=b({},e);return t.runtime="mediapipe",null==t.modelType&&(t.modelType=B.modelType),null==t.maxFaces&&(t.maxFaces=B.maxFaces),t}(e),[4,(n=new z(t)).initialize()];case 1:return i.sent(),[2,n]}}))}))}function A(e,t,n,i){var o=e.width,r=e.height,a=1,s=Math.cos(e.rotation),h=Math.sin(e.rotation),u=e.xCenter,c=e.yCenter,l=1/t,f=1/n,d=new Array(16);return d[0]=o*s*a*l,d[1]=-r*h*l,d[2]=0,d[3]=(-0.5*o*s*a+.5*r*h+u)*l,d[4]=o*h*a*f,d[5]=r*s*f,d[6]=0,d[7]=(-0.5*r*s-.5*o*h*a+c)*f,d[8]=0,d[9]=0,d[10]=o*l,d[11]=0,d[12]=0,d[13]=0,d[14]=0,d[15]=1,function(e){if(16!==e.length)throw new Error("Array length must be 16 but got ".concat(e.length));return [[e[0],e[1],e[2],e[3]],[e[4],e[5],e[6],e[7]],[e[8],e[9],e[10],e[11]],[e[12],e[13],e[14],e[15]]]}(d)}function F(e){return e instanceof tfjsCore.Tensor?{height:e.shape[0],width:e.shape[1]}:{height:e.height,width:e.width}}function E(e){return e instanceof tfjsCore.Tensor?e:tfjsCore.browser.fromPixels(e)}function R(e,t){tfjsCore.util.assert(0!==e.width,(function(){return "".concat(t," width cannot be 0.")})),tfjsCore.util.assert(0!==e.height,(function(){return "".concat(t," height cannot be 0.")}));}function L(e,t){var n=function(e,t,n,i){var o=t-e,r=i-n;var a=r/o;return {scale:a,offset:n-e*a}}(0,255,t[0],t[1]);return tfjsCore.tidy((function(){return tfjsCore.add(tfjsCore.mul(e,n.scale),n.offset)}))}function K(e,t,n){var i=t.outputTensorSize,r=t.keepAspectRatio,a=t.borderMode,l=t.outputTensorFloatRange,f=F(e),d=function(e,t){return {xCenter:.5*e.width,yCenter:.5*e.height,width:e.width,height:e.height,rotation:0}}(f),p=function(e,t,n){if(void 0===n&&(n=false),!n)return {top:0,left:0,right:0,bottom:0};var i=t.height,o=t.width;R(t,"targetSize"),R(e,"roi");var r,a,s=i/o,h=e.height/e.width,u=0,c=0;return s>h?(r=e.width,a=e.width*s,c=(1-h/s)/2):(r=e.height/s,a=e.height,u=(1-s/h)/2),e.width=r,e.height=a,{top:c,left:u,right:u,bottom:c}}(d,i,r),m=A(d,f.width,f.height),x=tfjsCore.tidy((function(){var t=E(e),n=tfjsCore.tensor2d(function(e,t,n){return R(n,"inputResolution"),[1/n.width*e[0][0]*t.width,1/n.height*e[0][1]*t.width,e[0][3]*t.width,1/n.width*e[1][0]*t.height,1/n.height*e[1][1]*t.height,e[1][3]*t.height,0,0]}(m,f,i),[1,8]),o="zero"===a?"constant":"nearest",r=tfjsCore.image.transform(tfjsCore.expandDims(tfjsCore.cast(t,"float32")),n,"bilinear",o,0,[i.height,i.width]);return null!=l?L(r,l):r}));return {imageTensor:x,padding:p,transformationMatrix:m}}function k(e){null==e.reduceBoxesInLowestLayer&&(e.reduceBoxesInLowestLayer=false),null==e.interpolatedScaleAspectRatio&&(e.interpolatedScaleAspectRatio=1),null==e.fixedAnchorSize&&(e.fixedAnchorSize=false);for(var t=[],n=0;n<e.numLayers;){for(var i=[],o=[],r=[],a=[],s=n;s<e.strides.length&&e.strides[s]===e.strides[n];){var h=P(e.minScale,e.maxScale,s,e.strides.length);if(0===s&&e.reduceBoxesInLowestLayer)r.push(1),r.push(2),r.push(.5),a.push(.1),a.push(h),a.push(h);else {for(var u=0;u<e.aspectRatios.length;++u)r.push(e.aspectRatios[u]),a.push(h);if(e.interpolatedScaleAspectRatio>0){var c=s===e.strides.length-1?1:P(e.minScale,e.maxScale,s+1,e.strides.length);a.push(Math.sqrt(h*c)),r.push(e.interpolatedScaleAspectRatio);}}s++;}for(var l=0;l<r.length;++l){var f=Math.sqrt(r[l]);i.push(a[l]/f),o.push(a[l]*f);}var d=0,p=0;if(e.featureMapHeight.length>0)d=e.featureMapHeight[n],p=e.featureMapWidth[n];else {var m=e.strides[n];d=Math.ceil(e.inputSizeHeight/m),p=Math.ceil(e.inputSizeWidth/m);}for(var x=0;x<d;++x)for(var g=0;g<p;++g)for(var y=0;y<i.length;++y){var v={xCenter:(g+e.anchorOffsetX)/p,yCenter:(x+e.anchorOffsetY)/d,width:0,height:0};e.fixedAnchorSize?(v.width=1,v.height=1):(v.width=o[y],v.height=i[y]),t.push(v);}n=s;}return t}function P(e,t,n,i){return 1===i?.5*(e+t):e+(t-e)*n/(i-1)}function V(e,t){var n=t[0],i=t[1];return [n*e[0]+i*e[1]+e[3],n*e[4]+i*e[5]+e[7]]}function H(e){return tfjsCore.tidy((function(){var t=function(e){return tfjsCore.tidy((function(){return [tfjsCore.slice(e,[0,0,0],[1,-1,1]),tfjsCore.slice(e,[0,0,1],[1,-1,-1])]}))}(e),n=t[0],i=t[1];return {boxes:tfjsCore.squeeze(i),logits:tfjsCore.squeeze(n)}}))}function U(e,t,n,i){return T(this,void 0,void 0,(function(){var i,o,r,a,u;return C(this,(function(c){switch(c.label){case 0:return e.sort((function(e,t){return Math.max.apply(Math,t.score)-Math.max.apply(Math,e.score)})),i=tfjsCore.tensor2d(e.map((function(e){return [e.locationData.relativeBoundingBox.yMin,e.locationData.relativeBoundingBox.xMin,e.locationData.relativeBoundingBox.yMax,e.locationData.relativeBoundingBox.xMax]}))),o=tfjsCore.tensor1d(e.map((function(e){return e.score[0]}))),[4,tfjsCore.image.nonMaxSuppressionAsync(i,o,t,n)];case 1:return [4,(r=c.sent()).array()];case 2:return a=c.sent(),u=e.filter((function(e,t){return a.indexOf(t)>-1})),tfjsCore.dispose([i,o,r]),[2,u]}}))}))}function j(e,t,n){return T(this,void 0,void 0,(function(){var i,s,h,u,c;return C(this,(function(p){switch(p.label){case 0:return i=e[0],s=e[1],h=function(e,t,n){return tfjsCore.tidy((function(){var i,o,s,h;n.reverseOutputOrder?(o=tfjsCore.squeeze(tfjsCore.slice(e,[0,n.boxCoordOffset+0],[-1,1])),i=tfjsCore.squeeze(tfjsCore.slice(e,[0,n.boxCoordOffset+1],[-1,1])),h=tfjsCore.squeeze(tfjsCore.slice(e,[0,n.boxCoordOffset+2],[-1,1])),s=tfjsCore.squeeze(tfjsCore.slice(e,[0,n.boxCoordOffset+3],[-1,1]))):(i=tfjsCore.squeeze(tfjsCore.slice(e,[0,n.boxCoordOffset+0],[-1,1])),o=tfjsCore.squeeze(tfjsCore.slice(e,[0,n.boxCoordOffset+1],[-1,1])),s=tfjsCore.squeeze(tfjsCore.slice(e,[0,n.boxCoordOffset+2],[-1,1])),h=tfjsCore.squeeze(tfjsCore.slice(e,[0,n.boxCoordOffset+3],[-1,1]))),o=tfjsCore.add(tfjsCore.mul(tfjsCore.div(o,n.xScale),t.w),t.x),i=tfjsCore.add(tfjsCore.mul(tfjsCore.div(i,n.yScale),t.h),t.y),n.applyExponentialOnBoxSize?(s=tfjsCore.mul(tfjsCore.exp(tfjsCore.div(s,n.hScale)),t.h),h=tfjsCore.mul(tfjsCore.exp(tfjsCore.div(h,n.wScale)),t.w)):(s=tfjsCore.mul(tfjsCore.div(s,n.hScale),t.h),h=tfjsCore.mul(tfjsCore.div(h,n.wScale),t.h));var u=tfjsCore.sub(i,tfjsCore.div(s,2)),c=tfjsCore.sub(o,tfjsCore.div(h,2)),d=tfjsCore.add(i,tfjsCore.div(s,2)),p=tfjsCore.add(o,tfjsCore.div(h,2)),w=tfjsCore.concat([tfjsCore.reshape(u,[n.numBoxes,1]),tfjsCore.reshape(c,[n.numBoxes,1]),tfjsCore.reshape(d,[n.numBoxes,1]),tfjsCore.reshape(p,[n.numBoxes,1])],1);if(n.numKeypoints)for(var M=0;M<n.numKeypoints;++M){var S=n.keypointCoordOffset+M*n.numValuesPerKeypoint,b=void 0,T=void 0;n.reverseOutputOrder?(b=tfjsCore.squeeze(tfjsCore.slice(e,[0,S],[-1,1])),T=tfjsCore.squeeze(tfjsCore.slice(e,[0,S+1],[-1,1]))):(T=tfjsCore.squeeze(tfjsCore.slice(e,[0,S],[-1,1])),b=tfjsCore.squeeze(tfjsCore.slice(e,[0,S+1],[-1,1])));var C=tfjsCore.add(tfjsCore.mul(tfjsCore.div(b,n.xScale),t.w),t.x),O=tfjsCore.add(tfjsCore.mul(tfjsCore.div(T,n.yScale),t.h),t.y);w=tfjsCore.concat([w,tfjsCore.reshape(C,[n.numBoxes,1]),tfjsCore.reshape(O,[n.numBoxes,1])],1);}return w}))}(s,t,n),u=tfjsCore.tidy((function(){var e=i;return n.sigmoidScore?(null!=n.scoreClippingThresh&&(e=tfjsCore.clipByValue(i,-n.scoreClippingThresh,n.scoreClippingThresh)),e=tfjsCore.sigmoid(e)):e})),[4,I(h,u,n)];case 1:return c=p.sent(),tfjsCore.dispose([h,u]),[2,c]}}))}))}function I(e,t,n){return T(this,void 0,void 0,(function(){var i,o,r,a,s,h,u,c,l,f,d,p;return C(this,(function(m){switch(m.label){case 0:return i=[],[4,e.data()];case 1:return o=m.sent(),[4,t.data()];case 2:for(r=m.sent(),a=0;a<n.numBoxes;++a)if(!(null!=n.minScoreThresh&&r[a]<n.minScoreThresh||(s=a*n.numCoords,h=_(o[s+0],o[s+1],o[s+2],o[s+3],r[a],n.flipVertically,a),(u=h.locationData.relativeBoundingBox).width<0||u.height<0))){if(n.numKeypoints>0)for((c=h.locationData).relativeKeypoints=[],l=n.numKeypoints*n.numValuesPerKeypoint,f=0;f<l;f+=n.numValuesPerKeypoint)d=s+n.keypointCoordOffset+f,p={x:o[d+0],y:n.flipVertically?1-o[d+1]:o[d+1]},c.relativeKeypoints.push(p);i.push(h);}return [2,i]}}))}))}function _(e,t,n,i,o,r,a){return {score:[o],ind:a,locationData:{relativeBoundingBox:{xMin:t,yMin:r?1-n:e,xMax:i,yMax:r?1-e:n,width:i-t,height:n-e}}}}var N={reduceBoxesInLowestLayer:false,interpolatedScaleAspectRatio:1,featureMapHeight:[],featureMapWidth:[],numLayers:4,minScale:.1484375,maxScale:.75,inputSizeHeight:128,inputSizeWidth:128,anchorOffsetX:.5,anchorOffsetY:.5,strides:[8,16,16,16],aspectRatios:[1],fixedAnchorSize:true},W={reduceBoxesInLowestLayer:false,interpolatedScaleAspectRatio:0,featureMapHeight:[],featureMapWidth:[],numLayers:1,minScale:.1484375,maxScale:.75,inputSizeHeight:192,inputSizeWidth:192,anchorOffsetX:.5,anchorOffsetY:.5,strides:[4],aspectRatios:[1],fixedAnchorSize:true},X={runtime:"tfjs",modelType:"short",maxFaces:1,detectorModelUrl:"https://tfhub.dev/mediapipe/tfjs-model/face_detection/short/1"},Y={applyExponentialOnBoxSize:false,flipVertically:false,ignoreClasses:[],numClasses:1,numBoxes:896,numCoords:16,boxCoordOffset:0,keypointCoordOffset:4,numKeypoints:6,numValuesPerKeypoint:2,sigmoidScore:true,scoreClippingThresh:100,reverseOutputOrder:true,xScale:128,yScale:128,hScale:128,wScale:128,minScoreThresh:.5},q={applyExponentialOnBoxSize:false,flipVertically:false,ignoreClasses:[],numClasses:1,numBoxes:2304,numCoords:16,boxCoordOffset:0,keypointCoordOffset:4,numKeypoints:6,numValuesPerKeypoint:2,sigmoidScore:true,scoreClippingThresh:100,reverseOutputOrder:true,xScale:192,yScale:192,hScale:192,wScale:192,minScoreThresh:.6},G=.3,$={outputTensorSize:{width:128,height:128},keepAspectRatio:true,outputTensorFloatRange:[-1,1],borderMode:"zero"},J={outputTensorSize:{width:192,height:192},keepAspectRatio:true,outputTensorFloatRange:[-1,1],borderMode:"zero"};var Q,Z=function(){function e(e,t,n){this.detectorModel=t,this.maxFaces=n,"full"===e?(this.imageToTensorConfig=J,this.tensorsToDetectionConfig=q,this.anchors=k(W)):(this.imageToTensorConfig=$,this.tensorsToDetectionConfig=Y,this.anchors=k(N));var i=tfjsCore.tensor1d(this.anchors.map((function(e){return e.width}))),o=tfjsCore.tensor1d(this.anchors.map((function(e){return e.height}))),r=tfjsCore.tensor1d(this.anchors.map((function(e){return e.xCenter}))),a=tfjsCore.tensor1d(this.anchors.map((function(e){return e.yCenter})));this.anchorTensor={x:r,y:a,w:i,h:o};}return e.prototype.dispose=function(){this.detectorModel.dispose(),tfjsCore.dispose([this.anchorTensor.x,this.anchorTensor.y,this.anchorTensor.w,this.anchorTensor.h]);},e.prototype.reset=function(){},e.prototype.detectFaces=function(e,t){return void 0===t&&(t=false),T(this,void 0,void 0,(function(){var n,i,r,a,s,l,p,m,x,g,y;return C(this,(function(v){switch(v.label){case 0:return null==e?(this.reset(),[2,[]]):(n=tfjsCore.tidy((function(){var n=tfjsCore.cast(E(e),"float32");if(t){n=tfjsCore.squeeze(tfjsCore.image.flipLeftRight(tfjsCore.expandDims(n,0)),[0]);}return n})),i=K(n,this.imageToTensorConfig),r=i.imageTensor,a=i.transformationMatrix,s=this.detectorModel.execute(r,"Identity:0"),l=H(s),p=l.boxes,[4,j([m=l.logits,p],this.anchorTensor,this.tensorsToDetectionConfig)]);case 1:return 0===(x=v.sent()).length?(tfjsCore.dispose([n,r,s,m,p]),[2,x]):[4,U(x,this.maxFaces,G)];case 2:return g=v.sent(),y=function(e,t){ void 0===e&&(e=[]);var n,i=(n=t,[].concat.apply([],n));return e.forEach((function(e){var t=e.locationData;t.relativeKeypoints.forEach((function(e){var t=V(i,[e.x,e.y]),n=t[0],o=t[1];e.x=n,e.y=o;}));var n=t.relativeBoundingBox,o=Number.MAX_VALUE,r=Number.MAX_VALUE,a=Number.MIN_VALUE,s=Number.MIN_VALUE;[[n.xMin,n.yMin],[n.xMin+n.width,n.yMin],[n.xMin+n.width,n.yMin+n.height],[n.xMin,n.yMin+n.height]].forEach((function(e){var t=V(i,e),n=t[0],h=t[1];o=Math.min(o,n),a=Math.max(a,n),r=Math.min(r,h),s=Math.max(s,h);})),t.relativeBoundingBox={xMin:o,xMax:a,yMin:r,yMax:s,width:a-o,height:s-r};})),e}(g,a),tfjsCore.dispose([n,r,s,m,p]),[2,y]}}))}))},e.prototype.estimateFaces=function(e,t){return T(this,void 0,void 0,(function(){var n,i;return C(this,(function(o){return n=F(e),i=!!t&&t.flipHorizontal,[2,this.detectFaces(e,i).then((function(e){return e.map((function(e){for(var t=e.locationData.relativeKeypoints.map((function(e,t){return b(b({},e),{x:e.x*n.width,y:e.y*n.height,name:O[t]})})),i=e.locationData.relativeBoundingBox,o=0,r=["width","xMax","xMin"];o<r.length;o++){i[r[o]]*=n.width;}for(var a=0,s=["height","yMax","yMin"];a<s.length;a++){i[s[a]]*=n.height;}return {keypoints:t,box:i}}))}))]}))}))},e}();function ee(e){return T(this,void 0,void 0,(function(){var t,n,i;return C(this,(function(o){switch(o.label){case 0:return t=function(e){if(null==e)return b({},X);var t=b({},e);null==t.modelType&&(t.modelType=X.modelType),null==t.maxFaces&&(t.maxFaces=X.maxFaces),null==t.detectorModelUrl&&("full"===t.modelType?t.detectorModelUrl="https://tfhub.dev/mediapipe/tfjs-model/face_detection/full/1":t.detectorModelUrl="https://tfhub.dev/mediapipe/tfjs-model/face_detection/short/1");return t}(e),n="string"==typeof t.detectorModelUrl&&t.detectorModelUrl.indexOf("https://tfhub.dev")>-1,[4,tfjsConverter.loadGraphModel(t.detectorModelUrl,{fromTFHub:n})];case 1:return i=o.sent(),[2,new Z(t.modelType,i,t.maxFaces)]}}))}))}function te(e,t){return T(this,void 0,void 0,(function(){var n,i;return C(this,(function(o){if(e===Q.MediaPipeFaceDetector){if(i=void 0,null!=(n=t)){if("tfjs"===n.runtime)return [2,ee(n)];if("mediapipe"===n.runtime)return [2,D(n)];i=n.runtime;}throw new Error("Expect modelConfig.runtime to be either 'tfjs' "+"or 'mediapipe', but got ".concat(i))}throw new Error("".concat(e," is not a supported model name."))}))}))}!function(e){e.MediaPipeFaceDetector="MediaPipeFaceDetector";}(Q||(Q={}));
+
 /**
  * Checks whether WebAssembly is supported in the current environment.
  */
@@ -580,7 +603,7 @@ class BlazeFaceFaceDetection extends FaceDetection {
         }
         try {
             await tfjsCore.ready();
-            this._model = await faceDetection.createDetector(faceDetection.SupportedModels.MediaPipeFaceDetector, {
+            this._model = await te(Q.MediaPipeFaceDetector, {
                 runtime: 'tfjs',
                 detectorModelUrl: this.faceModelUrl,
             });
@@ -1055,7 +1078,7 @@ function createRequestStreamGenerator(stream) {
     };
 }
 
-const VERSION = '3.4.7';
+const VERSION = '3.5.1';
 
 const BASE_USER_AGENT = `ui-react-liveness/${VERSION}`;
 const getLivenessUserAgent = () => {
@@ -1610,6 +1633,17 @@ const STATIC_VIDEO_CONSTRAINTS = {
 const CAMERA_ID_KEY = 'AmplifyLivenessCameraId';
 const DEFAULT_FACE_FIT_TIMEOUT = 7000;
 let responseStream;
+// Helper function to get selected device info
+const getSelectedDeviceInfo = (context) => {
+    const selected = context.videoAssociatedParams?.selectableDevices?.find((d) => d.deviceId === context.videoAssociatedParams?.selectedDeviceId);
+    return selected
+        ? {
+            deviceId: selected.deviceId,
+            groupId: selected.groupId,
+            label: selected.label ?? '',
+        }
+        : undefined;
+};
 const responseStreamActor = async (callback) => {
     try {
         const stream = await responseStream;
@@ -1670,6 +1704,9 @@ const responseStreamActor = async (callback) => {
         }
     }
 };
+function getLastSelectedCameraId() {
+    return localStorage.getItem(CAMERA_ID_KEY);
+}
 function setLastSelectedCameraId(deviceId) {
     localStorage.setItem(CAMERA_ID_KEY, deviceId);
 }
@@ -2222,6 +2259,9 @@ const livenessMachine = xstate.createMachine({
                 if (event.data.message.includes('15 fps')) {
                     errorState = LivenessErrorState.CAMERA_FRAMERATE_ERROR;
                 }
+                else if (event.data.message.includes(LivenessErrorState.DEFAULT_CAMERA_NOT_FOUND_ERROR)) {
+                    errorState = LivenessErrorState.DEFAULT_CAMERA_NOT_FOUND_ERROR;
+                }
                 else {
                     errorState = LivenessErrorState.CAMERA_ACCESS_ERROR;
                 }
@@ -2231,15 +2271,26 @@ const livenessMachine = xstate.createMachine({
                     state: errorState,
                     error: error,
                 };
-                context.componentProps.onError?.(livenessError);
+                context.componentProps.onError?.(livenessError, getSelectedDeviceInfo(context));
                 return errorState;
             },
         }),
         callMobileLandscapeWarningCallback: xstate.assign({
             errorState: () => LivenessErrorState.MOBILE_LANDSCAPE_ERROR,
         }),
+        getSelectedDeviceInfo: (context) => getSelectedDeviceInfo(context),
         callUserCancelCallback: (context) => {
-            context.componentProps.onUserCancel?.();
+            const { onUserCancel } = context.componentProps ?? {};
+            if (!onUserCancel) {
+                return;
+            }
+            try {
+                onUserCancel();
+            }
+            catch (callbackError) {
+                // eslint-disable-next-line no-console
+                console.error('Error in onUserCancel callback:', callbackError);
+            }
         },
         callUserTimeoutCallback: (context) => {
             const error = new Error(context.errorMessage ?? 'Client Timeout');
@@ -2248,14 +2299,14 @@ const livenessMachine = xstate.createMachine({
                 state: context.errorState,
                 error: error,
             };
-            context.componentProps.onError?.(livenessError);
+            context.componentProps.onError?.(livenessError, getSelectedDeviceInfo(context));
         },
         callErrorCallback: (context, event) => {
             const livenessError = {
                 state: context.errorState,
                 error: event.data?.error || event.data,
             };
-            context.componentProps.onError?.(livenessError);
+            context.componentProps.onError?.(livenessError, getSelectedDeviceInfo(context));
         },
         cleanUpResources: (context) => {
             const { freshnessColorEl } = context.freshnessColorAssociatedParams;
@@ -2366,10 +2417,42 @@ const livenessMachine = xstate.createMachine({
     services: {
         async checkVirtualCameraAndGetStream(context) {
             const { videoConstraints } = context.videoAssociatedParams;
-            // Get initial stream to enumerate devices with non-empty labels
-            const initialStream = await navigator.mediaDevices.getUserMedia({
-                video: { ...videoConstraints },
+            const { componentProps } = context;
+            let targetDeviceId;
+            let cameraNotFound = false;
+            if (componentProps?.config?.deviceId) {
+                targetDeviceId = componentProps.config.deviceId;
+            }
+            else {
+                targetDeviceId = getLastSelectedCameraId() ?? undefined;
+            }
+            const initialStream = await navigator.mediaDevices
+                .getUserMedia({
+                video: {
+                    ...videoConstraints,
+                    ...(targetDeviceId
+                        ? { deviceId: { exact: targetDeviceId } }
+                        : {}),
+                },
                 audio: false,
+            })
+                .catch((error) => {
+                if (error instanceof DOMException &&
+                    (error.name === 'NotFoundError' ||
+                        error.name === 'OverconstrainedError')) {
+                    // Mark camera as not found when a specific target device (either provided via props
+                    // or previously selected/saved as default) cannot be accessed.
+                    if (targetDeviceId && !cameraNotFound) {
+                        cameraNotFound = true;
+                    }
+                    return navigator.mediaDevices.getUserMedia({
+                        video: {
+                            ...videoConstraints,
+                        },
+                        audio: false,
+                    });
+                }
+                throw error;
             });
             const devices = await navigator.mediaDevices.enumerateDevices();
             const realVideoDevices = devices
@@ -2383,7 +2466,7 @@ const livenessMachine = xstate.createMachine({
                 .getTracks()
                 .filter((track) => {
                 const settings = track.getSettings();
-                return settings.frameRate >= 15;
+                return (settings.frameRate ?? 0) >= 15;
             });
             if (tracksWithMoreThan15Fps.length < 1) {
                 throw new Error('No camera found with more than 15 fps');
@@ -2402,11 +2485,16 @@ const livenessMachine = xstate.createMachine({
                 });
             }
             setLastSelectedCameraId(deviceId);
-            return {
+            const result = {
                 stream: realVideoDeviceStream,
                 selectedDeviceId: initialStreamDeviceId,
                 selectableDevices: realVideoDevices,
             };
+            // If a specific camera was requested but not found, trigger a specific error
+            if (cameraNotFound) {
+                throw new Error(LivenessErrorState.DEFAULT_CAMERA_NOT_FOUND_ERROR);
+            }
+            return result;
         },
         // eslint-disable-next-line @typescript-eslint/require-await
         async openLivenessStreamConnection(context) {
@@ -2659,9 +2747,20 @@ const livenessMachine = xstate.createMachine({
             livenessStreamProvider.dispatchStreamEvent({ type: 'streamStop' });
         },
         async getLiveness(context) {
-            const { onAnalysisComplete } = context.componentProps;
-            // Get liveness result
-            await onAnalysisComplete();
+            const { onAnalysisComplete } = context.componentProps ?? {};
+            if (!onAnalysisComplete) {
+                return;
+            }
+            try {
+                const deviceInfo = getSelectedDeviceInfo(context);
+                await onAnalysisComplete(deviceInfo);
+            }
+            catch (callbackError) {
+                // eslint-disable-next-line no-console
+                console.error('Error in onAnalysisComplete callback:', callbackError);
+                // Rethrow to allow the state machine to handle the error
+                throw callbackError;
+            }
         },
     },
 });

package/dist/styles.css

@@ -3565,6 +3565,23 @@ strong.amplify-text {
   gap: var(--amplify-space-medium);
 }
 
+[data-amplify-authenticator-passkeyprompt] .amplify-authenticator__passkey-success-icon {
+  font-size: var(--amplify-font-sizes-xxxl);
+  color: var(--amplify-colors-green-60, #34a853);
+}
+[data-amplify-authenticator-passkeyprompt] .amplify-authenticator__passkey-credential-item {
+  padding: var(--amplify-space-medium);
+  background-color: var(--amplify-colors-background-secondary);
+  border-radius: var(--amplify-radii-small);
+}
+[data-amplify-authenticator-passkeyprompt] .amplify-authenticator__passkey-error {
+  color: var(--amplify-colors-font-error);
+  margin-top: var(--amplify-space-small);
+}
+[data-amplify-authenticator-passkeyprompt] .amplify-authenticator__passkey-icon {
+  font-size: var(--amplify-components-authenticator-passkey-icon-size, 12rem);
+}
+
 .amplify-avatar {
   --avatar-color: var(--amplify-components-avatar-color);
   --avatar-background-color: var(--amplify-components-avatar-background-color);

package/dist/types/components/FaceLivenessDetector/service/machine/machine.d.ts

@@ -1,4 +1,5 @@
-import type { LivenessContext, LivenessEvent } from '../types';
+import type { LivenessContext, LivenessEvent, DeviceInfo } from '../types';
+export declare const getSelectedDeviceInfo: (context: LivenessContext) => DeviceInfo | undefined;
 export declare const livenessMachine: import("xstate").StateMachine<LivenessContext, any, LivenessEvent, {
     value: any;
     context: LivenessContext;

package/dist/types/components/FaceLivenessDetector/service/types/error.d.ts

@@ -12,5 +12,6 @@ export declare const LivenessErrorState: {
     readonly FACE_DISTANCE_ERROR: "FACE_DISTANCE_ERROR";
     readonly MOBILE_LANDSCAPE_ERROR: "MOBILE_LANDSCAPE_ERROR";
     readonly MULTIPLE_FACES_ERROR: "MULTIPLE_FACES_ERROR";
+    readonly DEFAULT_CAMERA_NOT_FOUND_ERROR: "DEFAULT_CAMERA_NOT_FOUND_ERROR";
 };
 export type ErrorState = keyof typeof LivenessErrorState;

package/dist/types/components/FaceLivenessDetector/service/types/liveness.d.ts

@@ -1,5 +1,10 @@
 import type { AwsCredentialProvider } from './credentials';
 import type { ErrorState } from './error';
+export interface DeviceInfo {
+    deviceId: string;
+    groupId: string;
+    label: string;
+}
 /**
  * The props for the FaceLivenessDetectorCore which allows for full configuration of auth
  */
@@ -11,8 +16,9 @@ export interface FaceLivenessDetectorCoreProps {
     /**
      * Callback that signals when the liveness session has completed analysis.
      * At this point a request can be made to GetFaceLivenessSessionResults.
+     * @param deviceInfo Information about the selected device
      */
-    onAnalysisComplete: () => Promise<void>;
+    onAnalysisComplete: (deviceInfo?: DeviceInfo) => Promise<void>;
     /**
      * The AWS region to stream the video to, for current regional support see the documentation here: FIXME LINK
      */
@@ -22,9 +28,11 @@ export interface FaceLivenessDetectorCoreProps {
      */
     onUserCancel?: () => void;
     /**
-     * Callback called when there is error occured on any step
+     * Callback called when there is an error on any step
+     * @param livenessError The error that occurred
+     * @param deviceInfo Information about the selected device, if available
      */
-    onError?: (livenessError: LivenessError) => void;
+    onError?: (livenessError: LivenessError, deviceInfo?: DeviceInfo) => void;
     /**
      * Optional parameter for the disabling the Start/Get Ready Screen, default: false
      */
@@ -44,6 +52,10 @@ export type FaceLivenessDetectorProps = Omit<FaceLivenessDetectorCoreProps, 'con
     config?: FaceLivenessDetectorConfig;
 };
 export interface FaceLivenessDetectorCoreConfig {
+    /**
+     * Optional device ID to pre-select a camera
+     */
+    deviceId?: string;
     /**
      * overrides the Wasm backend binary CDN path
      * default is https://cdn.jsdelivr.net/npm/@tensorflow/tfjs-backend-wasm@3.11.0/dist/.

package/dist/types/components/FaceLivenessDetector/service/types/machine.d.ts

@@ -66,7 +66,7 @@ export interface LivenessContext {
     shouldDisconnect: boolean | undefined;
     videoAssociatedParams: VideoAssociatedParams | undefined;
 }
-export type LivenessEventTypes = 'BEGIN' | 'CONNECTION_TIMEOUT' | 'START_RECORDING' | 'TIMEOUT' | 'ERROR' | 'CANCEL' | 'SET_SESSION_INFO' | 'DISCONNECT_EVENT' | 'SET_DOM_AND_CAMERA_DETAILS' | 'UPDATE_DEVICE_AND_STREAM' | 'SERVER_ERROR' | 'RUNTIME_ERROR' | 'RETRY_CAMERA_CHECK' | 'MOBILE_LANDSCAPE_WARNING';
+export type LivenessEventTypes = 'BEGIN' | 'CONNECTION_TIMEOUT' | 'START_RECORDING' | 'TIMEOUT' | 'ERROR' | 'CANCEL' | 'SET_SESSION_INFO' | 'DISCONNECT_EVENT' | 'SET_DOM_AND_CAMERA_DETAILS' | 'UPDATE_DEVICE_AND_STREAM' | 'CAMERA_NOT_FOUND' | 'SERVER_ERROR' | 'RUNTIME_ERROR' | 'RETRY_CAMERA_CHECK' | 'MOBILE_LANDSCAPE_WARNING';
 export type LivenessEventData = Record<PropertyKey, any>;
 export interface LivenessEvent {
     type: LivenessEventTypes;

package/dist/types/components/FaceLivenessDetector/service/utils/liveness.d.ts

@@ -76,6 +76,7 @@ export declare const LivenessErrorStateStringMap: {
     CAMERA_ACCESS_ERROR: string;
     MOBILE_LANDSCAPE_ERROR: string;
     FRESHNESS_TIMEOUT: string;
+    DEFAULT_CAMERA_NOT_FOUND_ERROR: string;
 };
 interface FillOverlayCanvasFractionalInput {
     overlayCanvas: HTMLCanvasElement;

package/dist/types/components/FaceLivenessDetector/shared/Hint.d.ts

@@ -1,7 +1,7 @@
 import * as React from 'react';
 import { IlluminationState, FaceMatchState } from '../service';
 import type { HintDisplayText } from '../displayText';
-export declare const selectErrorState: import("../hooks").LivenessSelectorFn<"CONNECTION_TIMEOUT" | "TIMEOUT" | "RUNTIME_ERROR" | "FRESHNESS_TIMEOUT" | "SERVER_ERROR" | "CAMERA_FRAMERATE_ERROR" | "CAMERA_ACCESS_ERROR" | "FACE_DISTANCE_ERROR" | "MOBILE_LANDSCAPE_ERROR" | "MULTIPLE_FACES_ERROR" | undefined>;
+export declare const selectErrorState: import("../hooks").LivenessSelectorFn<"CONNECTION_TIMEOUT" | "TIMEOUT" | "RUNTIME_ERROR" | "FRESHNESS_TIMEOUT" | "SERVER_ERROR" | "CAMERA_FRAMERATE_ERROR" | "CAMERA_ACCESS_ERROR" | "FACE_DISTANCE_ERROR" | "MOBILE_LANDSCAPE_ERROR" | "MULTIPLE_FACES_ERROR" | "DEFAULT_CAMERA_NOT_FOUND_ERROR" | undefined>;
 export declare const selectFaceMatchState: import("../hooks").LivenessSelectorFn<FaceMatchState | undefined>;
 export declare const selectIlluminationState: import("../hooks").LivenessSelectorFn<IlluminationState | undefined>;
 export declare const selectIsFaceFarEnoughBeforeRecording: import("../hooks").LivenessSelectorFn<boolean | undefined>;

package/dist/types/version.d.ts

@@ -1 +1 @@
-export declare const VERSION = "3.4.7";
+export declare const VERSION = "3.5.1";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aws-amplify/ui-react-liveness",
-  "version": "3.4.7",
+  "version": "3.5.1",
   "main": "dist/index.js",
   "module": "dist/esm/index.mjs",
   "exports": {
@@ -26,6 +26,7 @@
   },
   "files": [
     "dist",
+    "patches",
     "LICENSE"
   ],
   "scripts": {
@@ -35,7 +36,7 @@
     "clean": "rimraf dist node_modules",
     "dev": "yarn build:rollup --watch",
     "lint": "yarn typecheck && eslint .",
-    "prebuild": "rimraf dist",
+    "prebuild": "rimraf dist && patch-package",
     "size": "yarn size-limit",
     "test": "jest",
     "test:watch": "yarn test --watch",
@@ -43,22 +44,21 @@
   },
   "peerDependencies": {
     "@aws-amplify/core": "*",
-    "aws-amplify": "^6.14.3",
+    "aws-amplify": "^6.15.9",
     "react": "^16.14 || ^17 || ^18 || ^19",
     "react-dom": "^16.14 || ^17 || ^18 || ^19"
   },
   "dependencies": {
-    "@aws-amplify/ui": "6.12.1",
-    "@aws-amplify/ui-react": "6.13.1",
-    "@aws-sdk/client-rekognitionstreaming": "3.828.0",
-    "@aws-sdk/util-format-url": "3.609.0",
+    "@aws-amplify/ui": "6.14.0",
+    "@aws-amplify/ui-react": "6.14.0",
+    "@aws-sdk/client-rekognitionstreaming": "3.967.0",
+    "@aws-sdk/util-format-url": "3.965.0",
     "@smithy/eventstream-serde-browser": "^4.0.4",
     "@smithy/fetch-http-handler": "^5.0.4",
     "@smithy/protocol-http": "^3.0.3",
     "@smithy/signature-v4": "5.1.2",
     "@smithy/types": "^4.3.1",
-    "@mediapipe/face_detection": "~0.4.0",
-    "@tensorflow-models/face-detection": "1.0.2",
+    "@mediapipe/face_detection": "^0.4.1646425229",
     "@tensorflow/tfjs-backend-cpu": "4.11.0",
     "@tensorflow/tfjs-backend-wasm": "4.11.0",
     "@tensorflow/tfjs-converter": "4.11.0",
@@ -69,12 +69,14 @@
     "xstate": "^4.33.6"
   },
   "devDependencies": {
+    "@tensorflow-models/face-detection": "~1.0.3",
     "@types/jest-when": "^3.5.0",
     "blob-polyfill": "^7.0.20220408",
     "jest-canvas-mock": "^2.5.2",
     "jest-websocket-mock": "^2.4.1",
     "jest-when": "^3.5.1",
     "mock-socket": "^9.2.1",
+    "patch-package": "~8.0.1",
     "web-streams-polyfill": "^3.2.1"
   },
   "size-limit": [
@@ -82,7 +84,7 @@
       "name": "FaceLivenessDetector",
       "path": "dist/esm/index.mjs",
       "import": "{ FaceLivenessDetector }",
-      "limit": "225 kB"
+      "limit": "229 kB"
     }
   ]
-}
+}

package/patches/@tensorflow-models+face-detection+1.0.3.patch

@@ -0,0 +1,9 @@
+diff --git a/node_modules/@tensorflow-models/face-detection/dist/face-detection.esm.js b/node_modules/@tensorflow-models/face-detection/dist/face-detection.esm.js
+index 1eb84f8..05b860e 100644
+--- a/node_modules/@tensorflow-models/face-detection/dist/face-detection.esm.js
++++ b/node_modules/@tensorflow-models/face-detection/dist/face-detection.esm.js
+@@ -15,3 +15,3 @@
+     * =============================================================================
+     */
+-import{FaceDetection as e}from"@mediapipe/face_detection";import{Tensor as t,browser as n,util as i,tidy as o,add as r,mul as a,tensor2d as s,image as h,expandDims as u,cast as c,slice as l,squeeze as f,dispose as d,tensor1d as p,div as m,exp as x,sub as g,concat as y,reshape as v,clipByValue as w,sigmoid as M}from"@tensorflow/tfjs-core";import{loadGraphModel as S}from"@tensorflow/tfjs-converter";var b=function(){return b=Object.assign||function(e){for(var t,n=1,i=arguments.length;n<i;n++)for(var o in t=arguments[n])Object.prototype.hasOwnProperty.call(t,o)&&(e[o]=t[o]);return e},b.apply(this,arguments)};function T(e,t,n,i){return new(n||(n=Promise))((function(o,r){function a(e){try{h(i.next(e))}catch(e){r(e)}}function s(e){try{h(i.throw(e))}catch(e){r(e)}}function h(e){var t;e.done?o(e.value):(t=e.value,t instanceof n?t:new n((function(e){e(t)}))).then(a,s)}h((i=i.apply(e,t||[])).next())}))}function C(e,t){var n,i,o,r,a={label:0,sent:function(){if(1&o[0])throw o[1];return o[1]},trys:[],ops:[]};return r={next:s(0),throw:s(1),return:s(2)},"function"==typeof Symbol&&(r[Symbol.iterator]=function(){return this}),r;function s(r){return function(s){return function(r){if(n)throw new TypeError("Generator is already executing.");for(;a;)try{if(n=1,i&&(o=2&r[0]?i.return:r[0]?i.throw||((o=i.return)&&o.call(i),0):i.next)&&!(o=o.call(i,r[1])).done)return o;switch(i=0,o&&(r=[2&r[0],o.value]),r[0]){case 0:case 1:o=r;break;case 4:return a.label++,{value:r[1],done:!1};case 5:a.label++,i=r[1],r=[0];continue;case 7:r=a.ops.pop(),a.trys.pop();continue;default:if(!(o=a.trys,(o=o.length>0&&o[o.length-1])||6!==r[0]&&2!==r[0])){a=0;continue}if(3===r[0]&&(!o||r[1]>o[0]&&r[1]<o[3])){a.label=r[1];break}if(6===r[0]&&a.label<o[1]){a.label=o[1],o=r;break}if(o&&a.label<o[2]){a.label=o[2],a.ops.push(r);break}o[2]&&a.ops.pop(),a.trys.pop();continue}r=t.call(e,a)}catch(e){r=[6,e],i=0}finally{n=o=0}if(5&r[0])throw r[1];return{value:r[0]?r[1]:void 0,done:!0}}([r,s])}}}var O=["rightEye","leftEye","noseTip","mouthCenter","rightEarTragion","leftEarTragion"];var B={modelType:"short",runtime:"mediapipe",maxFaces:1};var z=function(){function i(t){var n=this;this.width=0,this.height=0,this.selfieMode=!1,this.faceDetectorSolution=new e({locateFile:function(e,n){if(t.solutionPath){var i=t.solutionPath.replace(/\/+$/,"");return"".concat(i,"/").concat(e)}return"".concat(n,"/").concat(e)}}),this.faceDetectorSolution.setOptions({selfieMode:this.selfieMode,model:t.modelType}),this.faceDetectorSolution.onResults((function(e){if(n.height=e.image.height,n.width=e.image.width,n.faces=[],null!==e.detections)for(var t=0,i=e.detections;t<i.length;t++){var o=i[t];n.faces.push(n.normalizedToAbsolute(o.landmarks,(r=o.boundingBox,a=void 0,s=void 0,h=void 0,a=r.xCenter-r.width/2,s=a+r.width,h=r.yCenter-r.height/2,{xMin:a,xMax:s,yMin:h,yMax:h+r.height,width:r.width,height:r.height})))}var r,a,s,h}))}return i.prototype.normalizedToAbsolute=function(e,t){var n=this;return{keypoints:e.map((function(e,t){return{x:e.x*n.width,y:e.y*n.height,name:O[t]}})),box:{xMin:t.xMin*this.width,yMin:t.yMin*this.height,xMax:t.xMax*this.width,yMax:t.yMax*this.height,width:t.width*this.width,height:t.height*this.height}}},i.prototype.estimateFaces=function(e,i){return T(this,void 0,void 0,(function(){var o,r;return C(this,(function(a){switch(a.label){case 0:return i&&i.flipHorizontal&&i.flipHorizontal!==this.selfieMode&&(this.selfieMode=i.flipHorizontal,this.faceDetectorSolution.setOptions({selfieMode:this.selfieMode})),e instanceof t?(r=ImageData.bind,[4,n.toPixels(e)]):[3,2];case 1:return o=new(r.apply(ImageData,[void 0,a.sent(),e.shape[1],e.shape[0]])),[3,3];case 2:o=e,a.label=3;case 3:return e=o,[4,this.faceDetectorSolution.send({image:e})];case 4:return a.sent(),[2,this.faces]}}))}))},i.prototype.dispose=function(){this.faceDetectorSolution.close()},i.prototype.reset=function(){this.faceDetectorSolution.reset(),this.width=0,this.height=0,this.faces=null,this.selfieMode=!1},i.prototype.initialize=function(){return this.faceDetectorSolution.initialize()},i}();function D(e){return T(this,void 0,void 0,(function(){var t,n;return C(this,(function(i){switch(i.label){case 0:return t=function(e){if(null==e)return b({},B);var t=b({},e);return t.runtime="mediapipe",null==t.modelType&&(t.modelType=B.modelType),null==t.maxFaces&&(t.maxFaces=B.maxFaces),t}(e),[4,(n=new z(t)).initialize()];case 1:return i.sent(),[2,n]}}))}))}function A(e,t,n,i){var o=e.width,r=e.height,a=i?-1:1,s=Math.cos(e.rotation),h=Math.sin(e.rotation),u=e.xCenter,c=e.yCenter,l=1/t,f=1/n,d=new Array(16);return d[0]=o*s*a*l,d[1]=-r*h*l,d[2]=0,d[3]=(-.5*o*s*a+.5*r*h+u)*l,d[4]=o*h*a*f,d[5]=r*s*f,d[6]=0,d[7]=(-.5*r*s-.5*o*h*a+c)*f,d[8]=0,d[9]=0,d[10]=o*l,d[11]=0,d[12]=0,d[13]=0,d[14]=0,d[15]=1,function(e){if(16!==e.length)throw new Error("Array length must be 16 but got ".concat(e.length));return[[e[0],e[1],e[2],e[3]],[e[4],e[5],e[6],e[7]],[e[8],e[9],e[10],e[11]],[e[12],e[13],e[14],e[15]]]}(d)}function F(e){return e instanceof t?{height:e.shape[0],width:e.shape[1]}:{height:e.height,width:e.width}}function E(e){return e instanceof t?e:n.fromPixels(e)}function R(e,t){i.assert(0!==e.width,(function(){return"".concat(t," width cannot be 0.")})),i.assert(0!==e.height,(function(){return"".concat(t," height cannot be 0.")}))}function L(e,t){var n=function(e,t,n,i){var o=t-e,r=i-n;if(0===o)throw new Error("Original min and max are both ".concat(e,", range cannot be 0."));var a=r/o;return{scale:a,offset:n-e*a}}(0,255,t[0],t[1]);return o((function(){return r(a(e,n.scale),n.offset)}))}function K(e,t,n){var i=t.outputTensorSize,r=t.keepAspectRatio,a=t.borderMode,l=t.outputTensorFloatRange,f=F(e),d=function(e,t){return t?{xCenter:t.xCenter*e.width,yCenter:t.yCenter*e.height,width:t.width*e.width,height:t.height*e.height,rotation:t.rotation}:{xCenter:.5*e.width,yCenter:.5*e.height,width:e.width,height:e.height,rotation:0}}(f,n),p=function(e,t,n){if(void 0===n&&(n=!1),!n)return{top:0,left:0,right:0,bottom:0};var i=t.height,o=t.width;R(t,"targetSize"),R(e,"roi");var r,a,s=i/o,h=e.height/e.width,u=0,c=0;return s>h?(r=e.width,a=e.width*s,c=(1-h/s)/2):(r=e.height/s,a=e.height,u=(1-s/h)/2),e.width=r,e.height=a,{top:c,left:u,right:u,bottom:c}}(d,i,r),m=A(d,f.width,f.height,!1),x=o((function(){var t=E(e),n=s(function(e,t,n){return R(n,"inputResolution"),[1/n.width*e[0][0]*t.width,1/n.height*e[0][1]*t.width,e[0][3]*t.width,1/n.width*e[1][0]*t.height,1/n.height*e[1][1]*t.height,e[1][3]*t.height,0,0]}(m,f,i),[1,8]),o="zero"===a?"constant":"nearest",r=h.transform(u(c(t,"float32")),n,"bilinear",o,0,[i.height,i.width]);return null!=l?L(r,l):r}));return{imageTensor:x,padding:p,transformationMatrix:m}}function k(e){null==e.reduceBoxesInLowestLayer&&(e.reduceBoxesInLowestLayer=!1),null==e.interpolatedScaleAspectRatio&&(e.interpolatedScaleAspectRatio=1),null==e.fixedAnchorSize&&(e.fixedAnchorSize=!1);for(var t=[],n=0;n<e.numLayers;){for(var i=[],o=[],r=[],a=[],s=n;s<e.strides.length&&e.strides[s]===e.strides[n];){var h=P(e.minScale,e.maxScale,s,e.strides.length);if(0===s&&e.reduceBoxesInLowestLayer)r.push(1),r.push(2),r.push(.5),a.push(.1),a.push(h),a.push(h);else{for(var u=0;u<e.aspectRatios.length;++u)r.push(e.aspectRatios[u]),a.push(h);if(e.interpolatedScaleAspectRatio>0){var c=s===e.strides.length-1?1:P(e.minScale,e.maxScale,s+1,e.strides.length);a.push(Math.sqrt(h*c)),r.push(e.interpolatedScaleAspectRatio)}}s++}for(var l=0;l<r.length;++l){var f=Math.sqrt(r[l]);i.push(a[l]/f),o.push(a[l]*f)}var d=0,p=0;if(e.featureMapHeight.length>0)d=e.featureMapHeight[n],p=e.featureMapWidth[n];else{var m=e.strides[n];d=Math.ceil(e.inputSizeHeight/m),p=Math.ceil(e.inputSizeWidth/m)}for(var x=0;x<d;++x)for(var g=0;g<p;++g)for(var y=0;y<i.length;++y){var v={xCenter:(g+e.anchorOffsetX)/p,yCenter:(x+e.anchorOffsetY)/d,width:0,height:0};e.fixedAnchorSize?(v.width=1,v.height=1):(v.width=o[y],v.height=i[y]),t.push(v)}n=s}return t}function P(e,t,n,i){return 1===i?.5*(e+t):e+(t-e)*n/(i-1)}function V(e,t){var n=t[0],i=t[1];return[n*e[0]+i*e[1]+e[3],n*e[4]+i*e[5]+e[7]]}function H(e){return o((function(){var t=function(e){return o((function(){return[l(e,[0,0,0],[1,-1,1]),l(e,[0,0,1],[1,-1,-1])]}))}(e),n=t[0],i=t[1];return{boxes:f(i),logits:f(n)}}))}function U(e,t,n,i){return T(this,void 0,void 0,(function(){var i,o,r,a,u;return C(this,(function(c){switch(c.label){case 0:return e.sort((function(e,t){return Math.max.apply(Math,t.score)-Math.max.apply(Math,e.score)})),i=s(e.map((function(e){return[e.locationData.relativeBoundingBox.yMin,e.locationData.relativeBoundingBox.xMin,e.locationData.relativeBoundingBox.yMax,e.locationData.relativeBoundingBox.xMax]}))),o=p(e.map((function(e){return e.score[0]}))),[4,h.nonMaxSuppressionAsync(i,o,t,n)];case 1:return[4,(r=c.sent()).array()];case 2:return a=c.sent(),u=e.filter((function(e,t){return a.indexOf(t)>-1})),d([i,o,r]),[2,u]}}))}))}function j(e,t,n){return T(this,void 0,void 0,(function(){var i,s,h,u,c;return C(this,(function(p){switch(p.label){case 0:return i=e[0],s=e[1],h=function(e,t,n){return o((function(){var i,o,s,h;n.reverseOutputOrder?(o=f(l(e,[0,n.boxCoordOffset+0],[-1,1])),i=f(l(e,[0,n.boxCoordOffset+1],[-1,1])),h=f(l(e,[0,n.boxCoordOffset+2],[-1,1])),s=f(l(e,[0,n.boxCoordOffset+3],[-1,1]))):(i=f(l(e,[0,n.boxCoordOffset+0],[-1,1])),o=f(l(e,[0,n.boxCoordOffset+1],[-1,1])),s=f(l(e,[0,n.boxCoordOffset+2],[-1,1])),h=f(l(e,[0,n.boxCoordOffset+3],[-1,1]))),o=r(a(m(o,n.xScale),t.w),t.x),i=r(a(m(i,n.yScale),t.h),t.y),n.applyExponentialOnBoxSize?(s=a(x(m(s,n.hScale)),t.h),h=a(x(m(h,n.wScale)),t.w)):(s=a(m(s,n.hScale),t.h),h=a(m(h,n.wScale),t.h));var u=g(i,m(s,2)),c=g(o,m(h,2)),d=r(i,m(s,2)),p=r(o,m(h,2)),w=y([v(u,[n.numBoxes,1]),v(c,[n.numBoxes,1]),v(d,[n.numBoxes,1]),v(p,[n.numBoxes,1])],1);if(n.numKeypoints)for(var M=0;M<n.numKeypoints;++M){var S=n.keypointCoordOffset+M*n.numValuesPerKeypoint,b=void 0,T=void 0;n.reverseOutputOrder?(b=f(l(e,[0,S],[-1,1])),T=f(l(e,[0,S+1],[-1,1]))):(T=f(l(e,[0,S],[-1,1])),b=f(l(e,[0,S+1],[-1,1])));var C=r(a(m(b,n.xScale),t.w),t.x),O=r(a(m(T,n.yScale),t.h),t.y);w=y([w,v(C,[n.numBoxes,1]),v(O,[n.numBoxes,1])],1)}return w}))}(s,t,n),u=o((function(){var e=i;return n.sigmoidScore?(null!=n.scoreClippingThresh&&(e=w(i,-n.scoreClippingThresh,n.scoreClippingThresh)),e=M(e)):e})),[4,I(h,u,n)];case 1:return c=p.sent(),d([h,u]),[2,c]}}))}))}function I(e,t,n){return T(this,void 0,void 0,(function(){var i,o,r,a,s,h,u,c,l,f,d,p;return C(this,(function(m){switch(m.label){case 0:return i=[],[4,e.data()];case 1:return o=m.sent(),[4,t.data()];case 2:for(r=m.sent(),a=0;a<n.numBoxes;++a)if(!(null!=n.minScoreThresh&&r[a]<n.minScoreThresh||(s=a*n.numCoords,h=_(o[s+0],o[s+1],o[s+2],o[s+3],r[a],n.flipVertically,a),(u=h.locationData.relativeBoundingBox).width<0||u.height<0))){if(n.numKeypoints>0)for((c=h.locationData).relativeKeypoints=[],l=n.numKeypoints*n.numValuesPerKeypoint,f=0;f<l;f+=n.numValuesPerKeypoint)d=s+n.keypointCoordOffset+f,p={x:o[d+0],y:n.flipVertically?1-o[d+1]:o[d+1]},c.relativeKeypoints.push(p);i.push(h)}return[2,i]}}))}))}function _(e,t,n,i,o,r,a){return{score:[o],ind:a,locationData:{relativeBoundingBox:{xMin:t,yMin:r?1-n:e,xMax:i,yMax:r?1-e:n,width:i-t,height:n-e}}}}var N={reduceBoxesInLowestLayer:!1,interpolatedScaleAspectRatio:1,featureMapHeight:[],featureMapWidth:[],numLayers:4,minScale:.1484375,maxScale:.75,inputSizeHeight:128,inputSizeWidth:128,anchorOffsetX:.5,anchorOffsetY:.5,strides:[8,16,16,16],aspectRatios:[1],fixedAnchorSize:!0},W={reduceBoxesInLowestLayer:!1,interpolatedScaleAspectRatio:0,featureMapHeight:[],featureMapWidth:[],numLayers:1,minScale:.1484375,maxScale:.75,inputSizeHeight:192,inputSizeWidth:192,anchorOffsetX:.5,anchorOffsetY:.5,strides:[4],aspectRatios:[1],fixedAnchorSize:!0},X={runtime:"tfjs",modelType:"short",maxFaces:1,detectorModelUrl:"https://tfhub.dev/mediapipe/tfjs-model/face_detection/short/1"},Y={applyExponentialOnBoxSize:!1,flipVertically:!1,ignoreClasses:[],numClasses:1,numBoxes:896,numCoords:16,boxCoordOffset:0,keypointCoordOffset:4,numKeypoints:6,numValuesPerKeypoint:2,sigmoidScore:!0,scoreClippingThresh:100,reverseOutputOrder:!0,xScale:128,yScale:128,hScale:128,wScale:128,minScoreThresh:.5},q={applyExponentialOnBoxSize:!1,flipVertically:!1,ignoreClasses:[],numClasses:1,numBoxes:2304,numCoords:16,boxCoordOffset:0,keypointCoordOffset:4,numKeypoints:6,numValuesPerKeypoint:2,sigmoidScore:!0,scoreClippingThresh:100,reverseOutputOrder:!0,xScale:192,yScale:192,hScale:192,wScale:192,minScoreThresh:.6},G=.3,$={outputTensorSize:{width:128,height:128},keepAspectRatio:!0,outputTensorFloatRange:[-1,1],borderMode:"zero"},J={outputTensorSize:{width:192,height:192},keepAspectRatio:!0,outputTensorFloatRange:[-1,1],borderMode:"zero"};var Q,Z=function(){function e(e,t,n){this.detectorModel=t,this.maxFaces=n,"full"===e?(this.imageToTensorConfig=J,this.tensorsToDetectionConfig=q,this.anchors=k(W)):(this.imageToTensorConfig=$,this.tensorsToDetectionConfig=Y,this.anchors=k(N));var i=p(this.anchors.map((function(e){return e.width}))),o=p(this.anchors.map((function(e){return e.height}))),r=p(this.anchors.map((function(e){return e.xCenter}))),a=p(this.anchors.map((function(e){return e.yCenter})));this.anchorTensor={x:r,y:a,w:i,h:o}}return e.prototype.dispose=function(){this.detectorModel.dispose(),d([this.anchorTensor.x,this.anchorTensor.y,this.anchorTensor.w,this.anchorTensor.h])},e.prototype.reset=function(){},e.prototype.detectFaces=function(e,t){return void 0===t&&(t=!1),T(this,void 0,void 0,(function(){var n,i,r,a,s,l,p,m,x,g,y;return C(this,(function(v){switch(v.label){case 0:return null==e?(this.reset(),[2,[]]):(n=o((function(){var n=c(E(e),"float32");if(t){n=f(h.flipLeftRight(u(n,0)),[0])}return n})),i=K(n,this.imageToTensorConfig),r=i.imageTensor,a=i.transformationMatrix,s=this.detectorModel.execute(r,"Identity:0"),l=H(s),p=l.boxes,[4,j([m=l.logits,p],this.anchorTensor,this.tensorsToDetectionConfig)]);case 1:return 0===(x=v.sent()).length?(d([n,r,s,m,p]),[2,x]):[4,U(x,this.maxFaces,G)];case 2:return g=v.sent(),y=function(e,t){void 0===e&&(e=[]);var n,i=(n=t,[].concat.apply([],n));return e.forEach((function(e){var t=e.locationData;t.relativeKeypoints.forEach((function(e){var t=V(i,[e.x,e.y]),n=t[0],o=t[1];e.x=n,e.y=o}));var n=t.relativeBoundingBox,o=Number.MAX_VALUE,r=Number.MAX_VALUE,a=Number.MIN_VALUE,s=Number.MIN_VALUE;[[n.xMin,n.yMin],[n.xMin+n.width,n.yMin],[n.xMin+n.width,n.yMin+n.height],[n.xMin,n.yMin+n.height]].forEach((function(e){var t=V(i,e),n=t[0],h=t[1];o=Math.min(o,n),a=Math.max(a,n),r=Math.min(r,h),s=Math.max(s,h)})),t.relativeBoundingBox={xMin:o,xMax:a,yMin:r,yMax:s,width:a-o,height:s-r}})),e}(g,a),d([n,r,s,m,p]),[2,y]}}))}))},e.prototype.estimateFaces=function(e,t){return T(this,void 0,void 0,(function(){var n,i;return C(this,(function(o){return n=F(e),i=!!t&&t.flipHorizontal,[2,this.detectFaces(e,i).then((function(e){return e.map((function(e){for(var t=e.locationData.relativeKeypoints.map((function(e,t){return b(b({},e),{x:e.x*n.width,y:e.y*n.height,name:O[t]})})),i=e.locationData.relativeBoundingBox,o=0,r=["width","xMax","xMin"];o<r.length;o++){i[r[o]]*=n.width}for(var a=0,s=["height","yMax","yMin"];a<s.length;a++){i[s[a]]*=n.height}return{keypoints:t,box:i}}))}))]}))}))},e}();function ee(e){return T(this,void 0,void 0,(function(){var t,n,i;return C(this,(function(o){switch(o.label){case 0:return t=function(e){if(null==e)return b({},X);var t=b({},e);null==t.modelType&&(t.modelType=X.modelType),null==t.maxFaces&&(t.maxFaces=X.maxFaces),null==t.detectorModelUrl&&("full"===t.modelType?t.detectorModelUrl="https://tfhub.dev/mediapipe/tfjs-model/face_detection/full/1":t.detectorModelUrl="https://tfhub.dev/mediapipe/tfjs-model/face_detection/short/1");return t}(e),n="string"==typeof t.detectorModelUrl&&t.detectorModelUrl.indexOf("https://tfhub.dev")>-1,[4,S(t.detectorModelUrl,{fromTFHub:n})];case 1:return i=o.sent(),[2,new Z(t.modelType,i,t.maxFaces)]}}))}))}function te(e,t){return T(this,void 0,void 0,(function(){var n,i;return C(this,(function(o){if(e===Q.MediaPipeFaceDetector){if(i=void 0,null!=(n=t)){if("tfjs"===n.runtime)return[2,ee(n)];if("mediapipe"===n.runtime)return[2,D(n)];i=n.runtime}throw new Error("Expect modelConfig.runtime to be either 'tfjs' "+"or 'mediapipe', but got ".concat(i))}throw new Error("".concat(e," is not a supported model name."))}))}))}!function(e){e.MediaPipeFaceDetector="MediaPipeFaceDetector"}(Q||(Q={}));export{z as MediaPipeFaceDetectorMediaPipe,Z as MediaPipeFaceDetectorTfjs,Q as SupportedModels,te as createDetector};
++import FaceDetection from"@mediapipe/face_detection";import{Tensor as t,browser as n,util as i,tidy as o,add as r,mul as a,tensor2d as s,image as h,expandDims as u,cast as c,slice as l,squeeze as f,dispose as d,tensor1d as p,div as m,exp as x,sub as g,concat as y,reshape as v,clipByValue as w,sigmoid as M}from"@tensorflow/tfjs-core";import{loadGraphModel as S}from"@tensorflow/tfjs-converter";var b=function(){return b=Object.assign||function(e){for(var t,n=1,i=arguments.length;n<i;n++)for(var o in t=arguments[n])Object.prototype.hasOwnProperty.call(t,o)&&(e[o]=t[o]);return e},b.apply(this,arguments)};function T(e,t,n,i){return new(n||(n=Promise))((function(o,r){function a(e){try{h(i.next(e))}catch(e){r(e)}}function s(e){try{h(i.throw(e))}catch(e){r(e)}}function h(e){var t;e.done?o(e.value):(t=e.value,t instanceof n?t:new n((function(e){e(t)}))).then(a,s)}h((i=i.apply(e,t||[])).next())}))}function C(e,t){var n,i,o,r,a={label:0,sent:function(){if(1&o[0])throw o[1];return o[1]},trys:[],ops:[]};return r={next:s(0),throw:s(1),return:s(2)},"function"==typeof Symbol&&(r[Symbol.iterator]=function(){return this}),r;function s(r){return function(s){return function(r){if(n)throw new TypeError("Generator is already executing.");for(;a;)try{if(n=1,i&&(o=2&r[0]?i.return:r[0]?i.throw||((o=i.return)&&o.call(i),0):i.next)&&!(o=o.call(i,r[1])).done)return o;switch(i=0,o&&(r=[2&r[0],o.value]),r[0]){case 0:case 1:o=r;break;case 4:return a.label++,{value:r[1],done:!1};case 5:a.label++,i=r[1],r=[0];continue;case 7:r=a.ops.pop(),a.trys.pop();continue;default:if(!(o=a.trys,(o=o.length>0&&o[o.length-1])||6!==r[0]&&2!==r[0])){a=0;continue}if(3===r[0]&&(!o||r[1]>o[0]&&r[1]<o[3])){a.label=r[1];break}if(6===r[0]&&a.label<o[1]){a.label=o[1],o=r;break}if(o&&a.label<o[2]){a.label=o[2],a.ops.push(r);break}o[2]&&a.ops.pop(),a.trys.pop();continue}r=t.call(e,a)}catch(e){r=[6,e],i=0}finally{n=o=0}if(5&r[0])throw r[1];return{value:r[0]?r[1]:void 0,done:!0}}([r,s])}}}var O=["rightEye","leftEye","noseTip","mouthCenter","rightEarTragion","leftEarTragion"];var B={modelType:"short",runtime:"mediapipe",maxFaces:1};var z=function(){function i(t){var n=this;this.width=0,this.height=0,this.selfieMode=!1,this.faceDetectorSolution=new FaceDetection({locateFile:function(e,n){if(t.solutionPath){var i=t.solutionPath.replace(/\/+$/,"");return"".concat(i,"/").concat(e)}return"".concat(n,"/").concat(e)}}),this.faceDetectorSolution.setOptions({selfieMode:this.selfieMode,model:t.modelType}),this.faceDetectorSolution.onResults((function(e){if(n.height=e.image.height,n.width=e.image.width,n.faces=[],null!==e.detections)for(var t=0,i=e.detections;t<i.length;t++){var o=i[t];n.faces.push(n.normalizedToAbsolute(o.landmarks,(r=o.boundingBox,a=void 0,s=void 0,h=void 0,a=r.xCenter-r.width/2,s=a+r.width,h=r.yCenter-r.height/2,{xMin:a,xMax:s,yMin:h,yMax:h+r.height,width:r.width,height:r.height})))}var r,a,s,h}))}return i.prototype.normalizedToAbsolute=function(e,t){var n=this;return{keypoints:e.map((function(e,t){return{x:e.x*n.width,y:e.y*n.height,name:O[t]}})),box:{xMin:t.xMin*this.width,yMin:t.yMin*this.height,xMax:t.xMax*this.width,yMax:t.yMax*this.height,width:t.width*this.width,height:t.height*this.height}}},i.prototype.estimateFaces=function(e,i){return T(this,void 0,void 0,(function(){var o,r;return C(this,(function(a){switch(a.label){case 0:return i&&i.flipHorizontal&&i.flipHorizontal!==this.selfieMode&&(this.selfieMode=i.flipHorizontal,this.faceDetectorSolution.setOptions({selfieMode:this.selfieMode})),e instanceof t?(r=ImageData.bind,[4,n.toPixels(e)]):[3,2];case 1:return o=new(r.apply(ImageData,[void 0,a.sent(),e.shape[1],e.shape[0]])),[3,3];case 2:o=e,a.label=3;case 3:return e=o,[4,this.faceDetectorSolution.send({image:e})];case 4:return a.sent(),[2,this.faces]}}))}))},i.prototype.dispose=function(){this.faceDetectorSolution.close()},i.prototype.reset=function(){this.faceDetectorSolution.reset(),this.width=0,this.height=0,this.faces=null,this.selfieMode=!1},i.prototype.initialize=function(){return this.faceDetectorSolution.initialize()},i}();function D(e){return T(this,void 0,void 0,(function(){var t,n;return C(this,(function(i){switch(i.label){case 0:return t=function(e){if(null==e)return b({},B);var t=b({},e);return t.runtime="mediapipe",null==t.modelType&&(t.modelType=B.modelType),null==t.maxFaces&&(t.maxFaces=B.maxFaces),t}(e),[4,(n=new z(t)).initialize()];case 1:return i.sent(),[2,n]}}))}))}function A(e,t,n,i){var o=e.width,r=e.height,a=i?-1:1,s=Math.cos(e.rotation),h=Math.sin(e.rotation),u=e.xCenter,c=e.yCenter,l=1/t,f=1/n,d=new Array(16);return d[0]=o*s*a*l,d[1]=-r*h*l,d[2]=0,d[3]=(-.5*o*s*a+.5*r*h+u)*l,d[4]=o*h*a*f,d[5]=r*s*f,d[6]=0,d[7]=(-.5*r*s-.5*o*h*a+c)*f,d[8]=0,d[9]=0,d[10]=o*l,d[11]=0,d[12]=0,d[13]=0,d[14]=0,d[15]=1,function(e){if(16!==e.length)throw new Error("Array length must be 16 but got ".concat(e.length));return[[e[0],e[1],e[2],e[3]],[e[4],e[5],e[6],e[7]],[e[8],e[9],e[10],e[11]],[e[12],e[13],e[14],e[15]]]}(d)}function F(e){return e instanceof t?{height:e.shape[0],width:e.shape[1]}:{height:e.height,width:e.width}}function E(e){return e instanceof t?e:n.fromPixels(e)}function R(e,t){i.assert(0!==e.width,(function(){return"".concat(t," width cannot be 0.")})),i.assert(0!==e.height,(function(){return"".concat(t," height cannot be 0.")}))}function L(e,t){var n=function(e,t,n,i){var o=t-e,r=i-n;if(0===o)throw new Error("Original min and max are both ".concat(e,", range cannot be 0."));var a=r/o;return{scale:a,offset:n-e*a}}(0,255,t[0],t[1]);return o((function(){return r(a(e,n.scale),n.offset)}))}function K(e,t,n){var i=t.outputTensorSize,r=t.keepAspectRatio,a=t.borderMode,l=t.outputTensorFloatRange,f=F(e),d=function(e,t){return t?{xCenter:t.xCenter*e.width,yCenter:t.yCenter*e.height,width:t.width*e.width,height:t.height*e.height,rotation:t.rotation}:{xCenter:.5*e.width,yCenter:.5*e.height,width:e.width,height:e.height,rotation:0}}(f,n),p=function(e,t,n){if(void 0===n&&(n=!1),!n)return{top:0,left:0,right:0,bottom:0};var i=t.height,o=t.width;R(t,"targetSize"),R(e,"roi");var r,a,s=i/o,h=e.height/e.width,u=0,c=0;return s>h?(r=e.width,a=e.width*s,c=(1-h/s)/2):(r=e.height/s,a=e.height,u=(1-s/h)/2),e.width=r,e.height=a,{top:c,left:u,right:u,bottom:c}}(d,i,r),m=A(d,f.width,f.height,!1),x=o((function(){var t=E(e),n=s(function(e,t,n){return R(n,"inputResolution"),[1/n.width*e[0][0]*t.width,1/n.height*e[0][1]*t.width,e[0][3]*t.width,1/n.width*e[1][0]*t.height,1/n.height*e[1][1]*t.height,e[1][3]*t.height,0,0]}(m,f,i),[1,8]),o="zero"===a?"constant":"nearest",r=h.transform(u(c(t,"float32")),n,"bilinear",o,0,[i.height,i.width]);return null!=l?L(r,l):r}));return{imageTensor:x,padding:p,transformationMatrix:m}}function k(e){null==e.reduceBoxesInLowestLayer&&(e.reduceBoxesInLowestLayer=!1),null==e.interpolatedScaleAspectRatio&&(e.interpolatedScaleAspectRatio=1),null==e.fixedAnchorSize&&(e.fixedAnchorSize=!1);for(var t=[],n=0;n<e.numLayers;){for(var i=[],o=[],r=[],a=[],s=n;s<e.strides.length&&e.strides[s]===e.strides[n];){var h=P(e.minScale,e.maxScale,s,e.strides.length);if(0===s&&e.reduceBoxesInLowestLayer)r.push(1),r.push(2),r.push(.5),a.push(.1),a.push(h),a.push(h);else{for(var u=0;u<e.aspectRatios.length;++u)r.push(e.aspectRatios[u]),a.push(h);if(e.interpolatedScaleAspectRatio>0){var c=s===e.strides.length-1?1:P(e.minScale,e.maxScale,s+1,e.strides.length);a.push(Math.sqrt(h*c)),r.push(e.interpolatedScaleAspectRatio)}}s++}for(var l=0;l<r.length;++l){var f=Math.sqrt(r[l]);i.push(a[l]/f),o.push(a[l]*f)}var d=0,p=0;if(e.featureMapHeight.length>0)d=e.featureMapHeight[n],p=e.featureMapWidth[n];else{var m=e.strides[n];d=Math.ceil(e.inputSizeHeight/m),p=Math.ceil(e.inputSizeWidth/m)}for(var x=0;x<d;++x)for(var g=0;g<p;++g)for(var y=0;y<i.length;++y){var v={xCenter:(g+e.anchorOffsetX)/p,yCenter:(x+e.anchorOffsetY)/d,width:0,height:0};e.fixedAnchorSize?(v.width=1,v.height=1):(v.width=o[y],v.height=i[y]),t.push(v)}n=s}return t}function P(e,t,n,i){return 1===i?.5*(e+t):e+(t-e)*n/(i-1)}function V(e,t){var n=t[0],i=t[1];return[n*e[0]+i*e[1]+e[3],n*e[4]+i*e[5]+e[7]]}function H(e){return o((function(){var t=function(e){return o((function(){return[l(e,[0,0,0],[1,-1,1]),l(e,[0,0,1],[1,-1,-1])]}))}(e),n=t[0],i=t[1];return{boxes:f(i),logits:f(n)}}))}function U(e,t,n,i){return T(this,void 0,void 0,(function(){var i,o,r,a,u;return C(this,(function(c){switch(c.label){case 0:return e.sort((function(e,t){return Math.max.apply(Math,t.score)-Math.max.apply(Math,e.score)})),i=s(e.map((function(e){return[e.locationData.relativeBoundingBox.yMin,e.locationData.relativeBoundingBox.xMin,e.locationData.relativeBoundingBox.yMax,e.locationData.relativeBoundingBox.xMax]}))),o=p(e.map((function(e){return e.score[0]}))),[4,h.nonMaxSuppressionAsync(i,o,t,n)];case 1:return[4,(r=c.sent()).array()];case 2:return a=c.sent(),u=e.filter((function(e,t){return a.indexOf(t)>-1})),d([i,o,r]),[2,u]}}))}))}function j(e,t,n){return T(this,void 0,void 0,(function(){var i,s,h,u,c;return C(this,(function(p){switch(p.label){case 0:return i=e[0],s=e[1],h=function(e,t,n){return o((function(){var i,o,s,h;n.reverseOutputOrder?(o=f(l(e,[0,n.boxCoordOffset+0],[-1,1])),i=f(l(e,[0,n.boxCoordOffset+1],[-1,1])),h=f(l(e,[0,n.boxCoordOffset+2],[-1,1])),s=f(l(e,[0,n.boxCoordOffset+3],[-1,1]))):(i=f(l(e,[0,n.boxCoordOffset+0],[-1,1])),o=f(l(e,[0,n.boxCoordOffset+1],[-1,1])),s=f(l(e,[0,n.boxCoordOffset+2],[-1,1])),h=f(l(e,[0,n.boxCoordOffset+3],[-1,1]))),o=r(a(m(o,n.xScale),t.w),t.x),i=r(a(m(i,n.yScale),t.h),t.y),n.applyExponentialOnBoxSize?(s=a(x(m(s,n.hScale)),t.h),h=a(x(m(h,n.wScale)),t.w)):(s=a(m(s,n.hScale),t.h),h=a(m(h,n.wScale),t.h));var u=g(i,m(s,2)),c=g(o,m(h,2)),d=r(i,m(s,2)),p=r(o,m(h,2)),w=y([v(u,[n.numBoxes,1]),v(c,[n.numBoxes,1]),v(d,[n.numBoxes,1]),v(p,[n.numBoxes,1])],1);if(n.numKeypoints)for(var M=0;M<n.numKeypoints;++M){var S=n.keypointCoordOffset+M*n.numValuesPerKeypoint,b=void 0,T=void 0;n.reverseOutputOrder?(b=f(l(e,[0,S],[-1,1])),T=f(l(e,[0,S+1],[-1,1]))):(T=f(l(e,[0,S],[-1,1])),b=f(l(e,[0,S+1],[-1,1])));var C=r(a(m(b,n.xScale),t.w),t.x),O=r(a(m(T,n.yScale),t.h),t.y);w=y([w,v(C,[n.numBoxes,1]),v(O,[n.numBoxes,1])],1)}return w}))}(s,t,n),u=o((function(){var e=i;return n.sigmoidScore?(null!=n.scoreClippingThresh&&(e=w(i,-n.scoreClippingThresh,n.scoreClippingThresh)),e=M(e)):e})),[4,I(h,u,n)];case 1:return c=p.sent(),d([h,u]),[2,c]}}))}))}function I(e,t,n){return T(this,void 0,void 0,(function(){var i,o,r,a,s,h,u,c,l,f,d,p;return C(this,(function(m){switch(m.label){case 0:return i=[],[4,e.data()];case 1:return o=m.sent(),[4,t.data()];case 2:for(r=m.sent(),a=0;a<n.numBoxes;++a)if(!(null!=n.minScoreThresh&&r[a]<n.minScoreThresh||(s=a*n.numCoords,h=_(o[s+0],o[s+1],o[s+2],o[s+3],r[a],n.flipVertically,a),(u=h.locationData.relativeBoundingBox).width<0||u.height<0))){if(n.numKeypoints>0)for((c=h.locationData).relativeKeypoints=[],l=n.numKeypoints*n.numValuesPerKeypoint,f=0;f<l;f+=n.numValuesPerKeypoint)d=s+n.keypointCoordOffset+f,p={x:o[d+0],y:n.flipVertically?1-o[d+1]:o[d+1]},c.relativeKeypoints.push(p);i.push(h)}return[2,i]}}))}))}function _(e,t,n,i,o,r,a){return{score:[o],ind:a,locationData:{relativeBoundingBox:{xMin:t,yMin:r?1-n:e,xMax:i,yMax:r?1-e:n,width:i-t,height:n-e}}}}var N={reduceBoxesInLowestLayer:!1,interpolatedScaleAspectRatio:1,featureMapHeight:[],featureMapWidth:[],numLayers:4,minScale:.1484375,maxScale:.75,inputSizeHeight:128,inputSizeWidth:128,anchorOffsetX:.5,anchorOffsetY:.5,strides:[8,16,16,16],aspectRatios:[1],fixedAnchorSize:!0},W={reduceBoxesInLowestLayer:!1,interpolatedScaleAspectRatio:0,featureMapHeight:[],featureMapWidth:[],numLayers:1,minScale:.1484375,maxScale:.75,inputSizeHeight:192,inputSizeWidth:192,anchorOffsetX:.5,anchorOffsetY:.5,strides:[4],aspectRatios:[1],fixedAnchorSize:!0},X={runtime:"tfjs",modelType:"short",maxFaces:1,detectorModelUrl:"https://tfhub.dev/mediapipe/tfjs-model/face_detection/short/1"},Y={applyExponentialOnBoxSize:!1,flipVertically:!1,ignoreClasses:[],numClasses:1,numBoxes:896,numCoords:16,boxCoordOffset:0,keypointCoordOffset:4,numKeypoints:6,numValuesPerKeypoint:2,sigmoidScore:!0,scoreClippingThresh:100,reverseOutputOrder:!0,xScale:128,yScale:128,hScale:128,wScale:128,minScoreThresh:.5},q={applyExponentialOnBoxSize:!1,flipVertically:!1,ignoreClasses:[],numClasses:1,numBoxes:2304,numCoords:16,boxCoordOffset:0,keypointCoordOffset:4,numKeypoints:6,numValuesPerKeypoint:2,sigmoidScore:!0,scoreClippingThresh:100,reverseOutputOrder:!0,xScale:192,yScale:192,hScale:192,wScale:192,minScoreThresh:.6},G=.3,$={outputTensorSize:{width:128,height:128},keepAspectRatio:!0,outputTensorFloatRange:[-1,1],borderMode:"zero"},J={outputTensorSize:{width:192,height:192},keepAspectRatio:!0,outputTensorFloatRange:[-1,1],borderMode:"zero"};var Q,Z=function(){function e(e,t,n){this.detectorModel=t,this.maxFaces=n,"full"===e?(this.imageToTensorConfig=J,this.tensorsToDetectionConfig=q,this.anchors=k(W)):(this.imageToTensorConfig=$,this.tensorsToDetectionConfig=Y,this.anchors=k(N));var i=p(this.anchors.map((function(e){return e.width}))),o=p(this.anchors.map((function(e){return e.height}))),r=p(this.anchors.map((function(e){return e.xCenter}))),a=p(this.anchors.map((function(e){return e.yCenter})));this.anchorTensor={x:r,y:a,w:i,h:o}}return e.prototype.dispose=function(){this.detectorModel.dispose(),d([this.anchorTensor.x,this.anchorTensor.y,this.anchorTensor.w,this.anchorTensor.h])},e.prototype.reset=function(){},e.prototype.detectFaces=function(e,t){return void 0===t&&(t=!1),T(this,void 0,void 0,(function(){var n,i,r,a,s,l,p,m,x,g,y;return C(this,(function(v){switch(v.label){case 0:return null==e?(this.reset(),[2,[]]):(n=o((function(){var n=c(E(e),"float32");if(t){n=f(h.flipLeftRight(u(n,0)),[0])}return n})),i=K(n,this.imageToTensorConfig),r=i.imageTensor,a=i.transformationMatrix,s=this.detectorModel.execute(r,"Identity:0"),l=H(s),p=l.boxes,[4,j([m=l.logits,p],this.anchorTensor,this.tensorsToDetectionConfig)]);case 1:return 0===(x=v.sent()).length?(d([n,r,s,m,p]),[2,x]):[4,U(x,this.maxFaces,G)];case 2:return g=v.sent(),y=function(e,t){void 0===e&&(e=[]);var n,i=(n=t,[].concat.apply([],n));return e.forEach((function(e){var t=e.locationData;t.relativeKeypoints.forEach((function(e){var t=V(i,[e.x,e.y]),n=t[0],o=t[1];e.x=n,e.y=o}));var n=t.relativeBoundingBox,o=Number.MAX_VALUE,r=Number.MAX_VALUE,a=Number.MIN_VALUE,s=Number.MIN_VALUE;[[n.xMin,n.yMin],[n.xMin+n.width,n.yMin],[n.xMin+n.width,n.yMin+n.height],[n.xMin,n.yMin+n.height]].forEach((function(e){var t=V(i,e),n=t[0],h=t[1];o=Math.min(o,n),a=Math.max(a,n),r=Math.min(r,h),s=Math.max(s,h)})),t.relativeBoundingBox={xMin:o,xMax:a,yMin:r,yMax:s,width:a-o,height:s-r}})),e}(g,a),d([n,r,s,m,p]),[2,y]}}))}))},e.prototype.estimateFaces=function(e,t){return T(this,void 0,void 0,(function(){var n,i;return C(this,(function(o){return n=F(e),i=!!t&&t.flipHorizontal,[2,this.detectFaces(e,i).then((function(e){return e.map((function(e){for(var t=e.locationData.relativeKeypoints.map((function(e,t){return b(b({},e),{x:e.x*n.width,y:e.y*n.height,name:O[t]})})),i=e.locationData.relativeBoundingBox,o=0,r=["width","xMax","xMin"];o<r.length;o++){i[r[o]]*=n.width}for(var a=0,s=["height","yMax","yMin"];a<s.length;a++){i[s[a]]*=n.height}return{keypoints:t,box:i}}))}))]}))}))},e}();function ee(e){return T(this,void 0,void 0,(function(){var t,n,i;return C(this,(function(o){switch(o.label){case 0:return t=function(e){if(null==e)return b({},X);var t=b({},e);null==t.modelType&&(t.modelType=X.modelType),null==t.maxFaces&&(t.maxFaces=X.maxFaces),null==t.detectorModelUrl&&("full"===t.modelType?t.detectorModelUrl="https://tfhub.dev/mediapipe/tfjs-model/face_detection/full/1":t.detectorModelUrl="https://tfhub.dev/mediapipe/tfjs-model/face_detection/short/1");return t}(e),n="string"==typeof t.detectorModelUrl&&t.detectorModelUrl.indexOf("https://tfhub.dev")>-1,[4,S(t.detectorModelUrl,{fromTFHub:n})];case 1:return i=o.sent(),[2,new Z(t.modelType,i,t.maxFaces)]}}))}))}function te(e,t){return T(this,void 0,void 0,(function(){var n,i;return C(this,(function(o){if(e===Q.MediaPipeFaceDetector){if(i=void 0,null!=(n=t)){if("tfjs"===n.runtime)return[2,ee(n)];if("mediapipe"===n.runtime)return[2,D(n)];i=n.runtime}throw new Error("Expect modelConfig.runtime to be either 'tfjs' "+"or 'mediapipe', but got ".concat(i))}throw new Error("".concat(e," is not a supported model name."))}))}))}!function(e){e.MediaPipeFaceDetector="MediaPipeFaceDetector"}(Q||(Q={}));export{z as MediaPipeFaceDetectorMediaPipe,Z as MediaPipeFaceDetectorTfjs,Q as SupportedModels,te as createDetector};