@stellarapp/tfjs-stellar 1.0.0 → 1.0.2

package/dist/src/models/u_net.js

@@ -0,0 +1,151 @@
+import * as tf from "@tensorflow/tfjs";
+export class UNetModel extends tf.Sequential {
+    constructor(args) {
+        const { filters, units, activation = units == 1 ? "sigmoid" : "softmax", depth, residual = false, batchNorm = false, inputShape = [null, null, 3], ...sequentialArgs } = args;
+        sequentialArgs.name = sequentialArgs.name ?? "unet_model";
+        super({
+            ...sequentialArgs,
+            // calling user should not modify the layers after instantiation
+            layers: [createUNet({ filters, units, activation, depth, residual, batchNorm, inputShape })]
+        });
+    }
+    summary(lineLength, positions, printFn) {
+        super.summary(lineLength, positions, printFn);
+        this.layers[0].summary(lineLength, positions, printFn);
+    }
+}
+export function createUNet({ filters, depth, units, activation, residual = false, batchNorm = false, inputShape = [null, null, 3] }) {
+    if (units < 1) {
+        throw Error(`createUNet: units should be >= 1, got ${units}`);
+    }
+    const [image_height, image_width] = inputShape;
+    const divisble_by = 2 ** depth;
+    if ((image_height != null && image_height % divisble_by != 0) ||
+        image_width != null && image_width % divisble_by != 0) {
+        throw Error(`createUNet: the input height and width must be divisible by 2^depth (${divisble_by})`);
+    }
+    const input = tf.input({ shape: inputShape });
+    const skip_connection = [];
+    let x = input;
+    // calculate the filter sizes for each level
+    const filter_sizes = Array.from({ length: depth }, (_, i) => filters * (2 ** i));
+    for (const filter_size of filter_sizes) {
+        const contraction = contractionBlock(x, filter_size, residual, batchNorm, `contraction-f${filter_size}`);
+        x = tf.layers.maxPooling2d({ poolSize: 2, strides: 2, name: `pool-f${filter_size}` }).apply(contraction);
+        skip_connection.push(contraction);
+    }
+    x = contractionBlock(x, filter_sizes.at(-1) * 2, residual, batchNorm, "bottleneck");
+    for (let i = filter_sizes.length - 1; i >= 0; i--) {
+        x = expansionBlock(x, skip_connection[i], filter_sizes[i], residual, batchNorm, `expansion-f${filter_sizes[i]}`);
+    }
+    const output = tf.layers.conv2d({
+        filters: units,
+        kernelSize: 1,
+        padding: "same",
+        activation: activation ?? (units == 1 ? "sigmoid" : "softmax"),
+        name: "output-conv"
+    }).apply(x);
+    return tf.model({ inputs: input, outputs: output, name: "u_net" });
+}
+export async function loadUNetModel(pathOrIOHandler, options) {
+    const model = await tf.loadLayersModel(pathOrIOHandler, options);
+    const unet = createUNet({ depth: 1, filters: 4, units: 1 }); // these are dummy args that are overwritten
+    const { name, ...rest } = model;
+    Object.assign(unet, rest);
+    return unet;
+}
+/**
+ * The contraction block of a U-Net
+ *
+ * Conv > BN > ReLU > Conv > BN + residual > ReLU
+ *
+ * TODO: for residual, change order to (BN > ReLU > Conv)x2 + residual
+ *
+ * @param x a previous layer's symbolic output
+ * @param filters the number of filters, usually half the previous expansion block's
+ * @param residual includes a residual connection
+ * @param batchNorm applies batch normalization before ReLU activation
+ * @param name a unique name for the contraction block
+ */
+function contractionBlock(x, filters, residual, batchNorm, name) {
+    const conv1 = tf.layers.conv2d({
+        filters,
+        kernelSize: 3,
+        padding: "same",
+        useBias: !batchNorm,
+        kernelInitializer: "heNormal",
+        name: `${name}-1-conv2d`
+    });
+    const relu1 = tf.layers.reLU({ name: `${name}-1-relu` });
+    const conv2 = tf.layers.conv2d({
+        filters,
+        kernelSize: 3,
+        padding: "same",
+        useBias: !batchNorm,
+        kernelInitializer: "heNormal",
+        name: `${name}-2-conv2d`
+    });
+    const relu2 = tf.layers.reLU({ name: `${name}-2-relu` });
+    let forward = conv1.apply(x);
+    if (batchNorm) {
+        forward = tf.layers.batchNormalization({ name: `${name}-1-batchnorm` }).apply(forward);
+    }
+    forward = relu1.apply(forward);
+    forward = conv2.apply(forward);
+    if (batchNorm) {
+        forward = tf.layers.batchNormalization({ name: `${name}-2-batchnorm` }).apply(forward);
+    }
+    if (residual) {
+        let residual_skip = x;
+        if (x.shape[x.shape.length - 1] != filters) {
+            // a 1x1 convolution on the input to ensure the residual connection's
+            // channels/filters dim matches the convolution output
+            residual_skip = tf.layers.conv2d({
+                filters,
+                kernelSize: 1,
+                padding: "same",
+                useBias: !batchNorm,
+                kernelInitializer: "heNormal",
+                name: `${name}-residual`
+            }).apply(x);
+        }
+        if (batchNorm) {
+            residual_skip = tf.layers.batchNormalization({
+                name: `${name}-residual-batchnorm`
+            }).apply(residual_skip);
+        }
+        forward = tf.layers.add().apply([
+            residual_skip,
+            forward
+        ]);
+    }
+    forward = relu2.apply(forward);
+    return forward;
+}
+/**
+ * The expansion block of a U-Net
+ *
+ * Upconv + skip > contraction block
+ *
+ * @param x a previous layer's symbolic output
+ * @param skip the corresponding contraction block's output (before pool), shape matches `x`
+ * @param filters the number of filters, usually half the previous expansion block's
+ * @param residual includes a residual connection
+ * @param batchNorm apply batch normalization, should be `false` when batch size is `1`
+ * @param name a unique name for the contraction block
+ */
+function expansionBlock(x, skip, filters, residual, batchNorm, name) {
+    const upconv = tf.layers.conv2dTranspose({
+        filters,
+        padding: "same",
+        kernelSize: 2,
+        strides: 2,
+        kernelInitializer: "heNormal",
+        name: `${name}-upconv`
+    });
+    const concat = tf.layers.concatenate({ axis: -1, name: `${name}-concat-upconv-skip` });
+    let forward = upconv.apply(x);
+    forward = concat.apply([forward, skip]);
+    return contractionBlock(forward, filters, residual, batchNorm, name);
+}
+//# sourceMappingURL=u_net.js.map

package/dist/src/models/u_net.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"u_net.js","sourceRoot":"","sources":["../../../src/models/u_net.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,kBAAkB,CAAC;AAuCvC,MAAM,OAAO,SAAU,SAAQ,EAAE,CAAC,UAAU;IAExC,YAAY,IAAmB;QAC3B,MAAM,EACF,OAAO,EACP,KAAK,EACL,UAAU,GAAG,KAAK,IAAI,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,SAAS,EAC/C,KAAK,EACL,QAAQ,GAAG,KAAK,EAChB,SAAS,GAAG,KAAK,EACjB,UAAU,GAAG,CAAC,IAAI,EAAE,IAAI,EAAE,CAAC,CAAC,EAC5B,GAAG,cAAc,EACpB,GAAG,IAAI,CAAC;QAET,cAAc,CAAC,IAAI,GAAG,cAAc,CAAC,IAAI,IAAI,YAAY,CAAC;QAE1D,KAAK,CAAC;YACF,GAAG,cAAc;YACjB,gEAAgE;YAChE,MAAM,EAAE,CAAC,UAAU,CAAC,EAAE,OAAO,EAAE,KAAK,EAAE,UAAU,EAAE,KAAK,EAAE,QAAQ,EAAE,SAAS,EAAE,UAAU,EAAE,CAAC,CAAC;SAC/F,CAAC,CAAC;IACP,CAAC;IAGQ,OAAO,CAAC,UAAmB,EAAE,SAAoB,EAAE,OAA2D;QACnH,KAAK,CAAC,OAAO,CAAC,UAAU,EAAE,SAAS,EAAE,OAAO,CAAC,CAAC;QAC7C,IAAI,CAAC,MAAM,CAAC,CAAC,CAAoB,CAAC,OAAO,CAAC,UAAU,EAAE,SAAS,EAAE,OAAO,CAAC,CAAC;IAC/E,CAAC;CACJ;AAGD,MAAM,UAAU,UAAU,CAAC,EAAE,OAAO,EAAE,KAAK,EAAE,KAAK,EAAE,UAAU,EAAE,QAAQ,GAAG,KAAK,EAAE,SAAS,GAAG,KAAK,EAAE,UAAU,GAAG,CAAC,IAAI,EAAE,IAAI,EAAE,CAAC,CAAC,EAAiB;IAC9I,IAAI,KAAK,GAAG,CAAC,EAAE,CAAC;QACZ,MAAM,KAAK,CAAC,yCAAyC,KAAK,EAAE,CAAC,CAAC;IAClE,CAAC;IAED,MAAM,CAAC,YAAY,EAAE,WAAW,CAAC,GAAG,UAAU,CAAC;IAC/C,MAAM,WAAW,GAAG,CAAC,IAAI,KAAK,CAAC;IAE/B,IAAI,CAAC,YAAY,IAAI,IAAI,IAAI,YAAY,GAAG,WAAW,IAAI,CAAC,CAAC;QACzD,WAAW,IAAI,IAAI,IAAI,WAAW,GAAG,WAAW,IAAI,CAAC,EAAE,CAAC;QACxD,MAAM,KAAK,CAAC,wEAAwE,WAAW,GAAG,CAAC,CAAA;IACvG,CAAC;IAED,MAAM,KAAK,GAAG,EAAE,CAAC,KAAK,CAAC,EAAE,KAAK,EAAE,UAAU,EAAE,CAAC,CAAC;IAE9C,MAAM,eAAe,GAAwB,EAAE,CAAC;IAEhD,IAAI,CAAC,GAAG,KAAK,CAAC;IAEd,4CAA4C;IAC5C,MAAM,YAAY,GAAG,KAAK,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,KAAK,EAAE,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,OAAO,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAEjF,KAAK,MAAM,WAAW,IAAI,YAAY,EAAE,CAAC;QACrC,MAAM,WAAW,GAAG,gBAAgB,CAAC,CAAC,EAAE,WAAW,EAAE,QAAQ,EAAE,SAAS,EAAE,gBAAgB,WAAW,EAAE,CAAC,CAAC;QAEzG,CAAC,GAAG,EAAE,CAAC,MAAM,CAAC,YAAY,CAAC,EAAE,QAAQ,EAAE,CAAC,EAAE,OAAO,EAAE,CAAC,EAAE,IAAI,EAAE,SAAS,WAAW,EAAE,EAAE,CAAC,CAAC,KAAK,CAAC,WAAW,CAAsB,CAAC;QAC9H,eAAe,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;IACtC,CAAC;IAED,CAAC,GAAG,gBAAgB,CAAC,CAAC,EAAE,YAAY,CAAC,EAAE,CAAC,CAAC,CAAC,CAAE,GAAG,CAAC,EAAE,QAAQ,EAAE,SAAS,EAAE,YAAY,CAAC,CAAC;IAErF,KAAK,IAAI,CAAC,GAAG,YAAY,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC,IAAI,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC;QAChD,CAAC,GAAG,cAAc,CAAC,CAAC,EAAE,eAAe,CAAC,CAAC,CAAC,EAAE,YAAY,CAAC,CAAC,CAAC,EAAE,QAAQ,EAAE,SAAS,EAAE,cAAc,YAAY,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;IACrH,CAAC;IAED,MAAM,MAAM,GAAG,EAAE,CAAC,MAAM,CAAC,MAAM,CAAC;QAC5B,OAAO,EAAE,KAAK;QACd,UAAU,EAAE,CAAC;QACb,OAAO,EAAE,MAAM;QACf,UAAU,EAAE,UAAU,IAAI,CAAC,KAAK,IAAI,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,SAAS,CAAC;QAC9D,IAAI,EAAE,aAAa;KACtB,CAAC,CAAC,KAAK,CAAC,CAAC,CAAsB,CAAC;IAEjC,OAAO,EAAE,CAAC,KAAK,CAAC,EAAE,MAAM,EAAE,KAAK,EAAE,OAAO,EAAE,MAAM,EAAE,IAAI,EAAE,OAAO,EAAE,CAAC,CAAC;AACvE,CAAC;AAGD,MAAM,CAAC,KAAK,UAAU,aAAa,CAAC,eAAyC,EAAE,OAA2B;IACtG,MAAM,KAAK,GAAG,MAAM,EAAE,CAAC,eAAe,CAAC,eAAe,EAAE,OAAO,CAAC,CAAC;IACjE,MAAM,IAAI,GAAG,UAAU,CAAC,EAAE,KAAK,EAAE,CAAC,EAAE,OAAO,EAAE,CAAC,EAAE,KAAK,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,4CAA4C;IACzG,MAAM,EAAE,IAAI,EAAE,GAAG,IAAI,EAAE,GAAG,KAAK,CAAC;IAChC,MAAM,CAAC,MAAM,CAAC,IAAI,EAAE,IAAI,CAAC,CAAC;IAE1B,OAAO,IAAI,CAAC;AAChB,CAAC;AAGD;;;;;;;;;;;;GAYG;AACH,SAAS,gBAAgB,CAAC,CAAoB,EAAE,OAAe,EAAE,QAAiB,EAAE,SAAkB,EAAE,IAAY;IAEhH,MAAM,KAAK,GAAG,EAAE,CAAC,MAAM,CAAC,MAAM,CAAC;QAC3B,OAAO;QACP,UAAU,EAAE,CAAC;QACb,OAAO,EAAE,MAAM;QACf,OAAO,EAAE,CAAC,SAAS;QACnB,iBAAiB,EAAE,UAAU;QAC7B,IAAI,EAAE,GAAG,IAAI,WAAW;KAC3B,CAAC,CAAC;IACH,MAAM,KAAK,GAAG,EAAE,CAAC,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,GAAG,IAAI,SAAS,EAAE,CAAC,CAAC;IAEzD,MAAM,KAAK,GAAG,EAAE,CAAC,MAAM,CAAC,MAAM,CAAC;QAC3B,OAAO;QACP,UAAU,EAAE,CAAC;QACb,OAAO,EAAE,MAAM;QACf,OAAO,EAAE,CAAC,SAAS;QACnB,iBAAiB,EAAE,UAAU;QAC7B,IAAI,EAAE,GAAG,IAAI,WAAW;KAC3B,CAAC,CAAC;IACH,MAAM,KAAK,GAAG,EAAE,CAAC,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,GAAG,IAAI,SAAS,EAAE,CAAC,CAAC;IAEzD,IAAI,OAAO,GAAG,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;IAE7B,IAAI,SAAS,EAAE,CAAC;QACZ,OAAO,GAAG,EAAE,CAAC,MAAM,CAAC,kBAAkB,CAAC,EAAE,IAAI,EAAE,GAAG,IAAI,cAAc,EAAE,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;IAC3F,CAAC;IAED,OAAO,GAAG,KAAK,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;IAE/B,OAAO,GAAG,KAAK,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;IAE/B,IAAI,SAAS,EAAE,CAAC;QACZ,OAAO,GAAG,EAAE,CAAC,MAAM,CAAC,kBAAkB,CAAC,EAAE,IAAI,EAAE,GAAG,IAAI,cAAc,EAAE,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;IAC3F,CAAC;IAED,IAAI,QAAQ,EAAE,CAAC;QACX,IAAI,aAAa,GAAG,CAAC,CAAC;QAEtB,IAAI,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,IAAI,OAAO,EAAE,CAAC;YACzC,qEAAqE;YACrE,sDAAsD;YACtD,aAAa,GAAG,EAAE,CAAC,MAAM,CAAC,MAAM,CAAC;gBAC7B,OAAO;gBACP,UAAU,EAAE,CAAC;gBACb,OAAO,EAAE,MAAM;gBACf,OAAO,EAAE,CAAC,SAAS;gBACnB,iBAAiB,EAAE,UAAU;gBAC7B,IAAI,EAAE,GAAG,IAAI,WAAW;aAC3B,CAAC,CAAC,KAAK,CAAC,CAAC,CAAsB,CAAC;QACrC,CAAC;QAED,IAAI,SAAS,EAAE,CAAC;YACZ,aAAa,GAAG,EAAE,CAAC,MAAM,CAAC,kBAAkB,CAAC;gBACzC,IAAI,EAAE,GAAG,IAAI,qBAAqB;aACrC,CAAC,CAAC,KAAK,CAAC,aAAa,CAAsB,CAAC;QACjD,CAAC;QAED,OAAO,GAAG,EAAE,CAAC,MAAM,CAAC,GAAG,EAAE,CAAC,KAAK,CAAC;YAC5B,aAAkC;YAClC,OAA4B;SAC/B,CAAC,CAAA;IACN,CAAC;IAED,OAAO,GAAG,KAAK,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;IAE/B,OAAO,OAA4B,CAAC;AACxC,CAAC;AAGD;;;;;;;;;;;GAWG;AACH,SAAS,cAAc,CAAC,CAAoB,EAAE,IAAuB,EAAE,OAAe,EAAE,QAAiB,EAAE,SAAkB,EAAE,IAAY;IAEvI,MAAM,MAAM,GAAG,EAAE,CAAC,MAAM,CAAC,eAAe,CAAC;QACrC,OAAO;QACP,OAAO,EAAE,MAAM;QACf,UAAU,EAAE,CAAC;QACb,OAAO,EAAE,CAAC;QACV,iBAAiB,EAAE,UAAU;QAC7B,IAAI,EAAE,GAAG,IAAI,SAAS;KACzB,CAAC,CAAC;IAEH,MAAM,MAAM,GAAG,EAAE,CAAC,MAAM,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,CAAC,CAAC,EAAE,IAAI,EAAE,GAAG,IAAI,qBAAqB,EAAE,CAAC,CAAC;IAEvF,IAAI,OAAO,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAsB,CAAC;IACnD,OAAO,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,OAAO,EAAE,IAAI,CAAC,CAAsB,CAAC;IAE7D,OAAO,gBAAgB,CAAC,OAAO,EAAE,OAAO,EAAE,QAAQ,EAAE,SAAS,EAAE,IAAI,CAAC,CAAC;AACzE,CAAC"}

package/{src/tfjs_types.ts → dist/src/tfjs_types.d.ts}

@@ -1,15 +1,10 @@
 import type { Tensor } from "@tensorflow/tfjs";
-
-
 export declare abstract class LazyIterator<T> {
     abstract next(): Promise<IteratorResult<T>>;
 }
-
-
 export declare abstract class Dataset<T> {
     abstract iterator(): Promise<LazyIterator<T>>;
     size: number;
 }
-
-
 export type LossOrMetricFn = (yTrue: Tensor, yPred: Tensor) => Tensor;
+//# sourceMappingURL=tfjs_types.d.ts.map

package/dist/src/tfjs_types.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"tfjs_types.d.ts","sourceRoot":"","sources":["../../src/tfjs_types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,EAAE,MAAM,kBAAkB,CAAC;AAG/C,MAAM,CAAC,OAAO,CAAC,QAAQ,OAAO,YAAY,CAAC,CAAC;IACxC,QAAQ,CAAC,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC;CAC9C;AAGD,MAAM,CAAC,OAAO,CAAC,QAAQ,OAAO,OAAO,CAAC,CAAC;IACnC,QAAQ,CAAC,QAAQ,IAAI,OAAO,CAAC,YAAY,CAAC,CAAC,CAAC,CAAC;IAC7C,IAAI,EAAE,MAAM,CAAC;CAChB;AAGD,MAAM,MAAM,cAAc,GAAG,CAAC,KAAK,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,KAAK,MAAM,CAAC"}

package/dist/src/tfjs_types.js

@@ -0,0 +1,2 @@
+export {};
+//# sourceMappingURL=tfjs_types.js.map

package/dist/src/tfjs_types.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"tfjs_types.js","sourceRoot":"","sources":["../../src/tfjs_types.ts"],"names":[],"mappings":""}

package/dist/src/utils.d.ts

@@ -0,0 +1,28 @@
+import * as tf from "@tensorflow/tfjs";
+/**
+ * Calculate the desired scaled image's height and width. The shortest edge will
+ * be scaled to match its corresponding target shape's edge. The longer
+ * edge might end up larger than intended.
+ *
+ * @param image_shape the `[height, width]` of the image
+ * @param target_shape the intended `[height, width]` of the final scaled image
+ */
+export declare function getScaleShape(image_shape: tf.Shape, target_shape: [number, number]): [scaled_height: number, scaled_width: number];
+/**
+ * Calculate the starting point for a crop (slice) operation
+ * on an image tensor with the shape `[height, width, channels]`.
+ *
+ * @param image_shape the `[height, width]` of the image
+ * @param target_shape the intended `[height, width]` of the final cropped image
+ */
+export declare function getRandomCropStart(image_shape: [height: number, width: number], target_shape: [height: number, width: number]): [number, number, number];
+/**
+ * Calculate the height and width padding such that the image is
+ * divisible by 2^depth.
+ *
+ * In U-Net image segmentation, the contraction and concatenate
+ * operations requires  the input image's height and width
+ * dimensions to be divisible by 2^depth.
+ */
+export declare function getPaddingForSegmentation(image: tf.Tensor3D, depth: number): [height: number, width: number];
+//# sourceMappingURL=utils.d.ts.map

package/dist/src/utils.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"utils.d.ts","sourceRoot":"","sources":["../../src/utils.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,kBAAkB,CAAC;AAGvC;;;;;;;GAOG;AACH,wBAAgB,aAAa,CAAC,WAAW,EAAE,EAAE,CAAC,KAAK,EAAE,YAAY,EAAE,CAAC,MAAM,EAAE,MAAM,CAAC,GAAG,CAAC,aAAa,EAAE,MAAM,EAAE,YAAY,EAAE,MAAM,CAAC,CAelI;AAGD;;;;;;GAMG;AACH,wBAAgB,kBAAkB,CAC9B,WAAW,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,CAAC,EAC5C,YAAY,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,CAAC,GAC9C,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,CAiB1B;AAGD;;;;;;;GAOG;AACH,wBAAgB,yBAAyB,CAAC,KAAK,EAAE,EAAE,CAAC,QAAQ,EAAE,KAAK,EAAE,MAAM,GAAG,CAAC,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,CAAC,CAS5G"}

package/{src/utils.ts → dist/src/utils.js}

@@ -1,18 +1,14 @@
-import * as tf from "@tensorflow/tfjs";
-
-
 /**
  * Calculate the desired scaled image's height and width. The shortest edge will
  * be scaled to match its corresponding target shape's edge. The longer
  * edge might end up larger than intended.
- * 
+ *
  * @param image_shape the `[height, width]` of the image
  * @param target_shape the intended `[height, width]` of the final scaled image
  */
-export function getScaleShape(image_shape: tf.Shape, target_shape: [number, number]): [scaled_height: number, scaled_width: number] {
-    const [img_height, img_width] = image_shape as [number, number, number];
+export function getScaleShape(image_shape, target_shape) {
+    const [img_height, img_width] = image_shape;
     const [target_height, target_width] = target_shape;
-
     // scale based on whichever target_edge / original_edge is largest,
     // we need the following to be true (1)
     // height * scale >= target_height
@@ -25,27 +21,20 @@ export function getScaleShape(image_shape: tf.Shape, target_shape: [number, numb
     const scale_factor = Math.max(target_height / img_height, target_width / img_width);
     return [Math.round(img_height * scale_factor), Math.round(img_width * scale_factor)];
 }
-
-
 /**
  * Calculate the starting point for a crop (slice) operation
  * on an image tensor with the shape `[height, width, channels]`.
- * 
+ *
  * @param image_shape the `[height, width]` of the image
  * @param target_shape the intended `[height, width]` of the final cropped image
  */
-export function getRandomCropStart(
-    image_shape: [height: number, width: number],
-    target_shape: [height: number, width: number]
-): [number, number, number] {
+export function getRandomCropStart(image_shape, target_shape) {
     const [img_height, img_width] = image_shape;
     const [crop_x, crop_y] = target_shape;
-
     if (img_height < crop_x || img_width < crop_y) {
         throw Error(`getRandomCropShape: cannot crop with a size that's bigger than,` +
             ` the image. Original [${img_height}, ${img_width}], crop [${crop_x}, ${crop_y}].`);
     }
-
     // there's a +1 because Math.random()'s range is [0, 1), excluding 1,
     // hence +1 to ensure the full range of possible crop starting points
     return [
@@ -53,34 +42,22 @@ export function getRandomCropStart(
         Math.floor(Math.random() * (img_height - crop_x + 1)),
         Math.floor(Math.random() * (img_width - crop_y + 1)),
         0 // not cropping channels, so it starts at the first index
-    ]
+    ];
 }
-
-
 /**
  * Calculate the height and width padding such that the image is
  * divisible by 2^depth.
- * 
+ *
  * In U-Net image segmentation, the contraction and concatenate
  * operations requires  the input image's height and width
  * dimensions to be divisible by 2^depth.
  */
-export function getPaddingForSegmentation(image: tf.Tensor3D, depth: number): [height: number, width: number] {
+export function getPaddingForSegmentation(image, depth) {
     const divisible = Math.pow(2, depth);
-
     const [height, width] = image.shape;
-
     return [
         (Math.ceil(height / divisible)) * divisible - height,
         (Math.ceil(width / divisible)) * divisible - width,
-    ]
-}
-
-
-export function generateCausalAttentionMask(query_seq_length: number, key_seq_length: number) {
-    return tf.tidy(() => {
-        return tf.linalg.bandPart(tf.ones([query_seq_length, key_seq_length]), -1, 0)
-            .sub(1)
-            .mul(1e7);
-    })
+    ];
 }
+//# sourceMappingURL=utils.js.map

package/dist/src/utils.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"utils.js","sourceRoot":"","sources":["../../src/utils.ts"],"names":[],"mappings":"AAGA;;;;;;;GAOG;AACH,MAAM,UAAU,aAAa,CAAC,WAAqB,EAAE,YAA8B;IAC/E,MAAM,CAAC,UAAU,EAAE,SAAS,CAAC,GAAG,WAAuC,CAAC;IACxE,MAAM,CAAC,aAAa,EAAE,YAAY,CAAC,GAAG,YAAY,CAAC;IAEnD,mEAAmE;IACnE,uCAAuC;IACvC,kCAAkC;IAClC,iCAAiC;IACjC,mDAAmD;IACnD,kCAAkC;IAClC,gCAAgC;IAChC,0FAA0F;IAC1F,oEAAoE;IACpE,MAAM,YAAY,GAAG,IAAI,CAAC,GAAG,CAAC,aAAa,GAAG,UAAU,EAAE,YAAY,GAAG,SAAS,CAAC,CAAC;IACpF,OAAO,CAAC,IAAI,CAAC,KAAK,CAAC,UAAU,GAAG,YAAY,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,SAAS,GAAG,YAAY,CAAC,CAAC,CAAC;AACzF,CAAC;AAGD;;;;;;GAMG;AACH,MAAM,UAAU,kBAAkB,CAC9B,WAA4C,EAC5C,YAA6C;IAE7C,MAAM,CAAC,UAAU,EAAE,SAAS,CAAC,GAAG,WAAW,CAAC;IAC5C,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,GAAG,YAAY,CAAC;IAEtC,IAAI,UAAU,GAAG,MAAM,IAAI,SAAS,GAAG,MAAM,EAAE,CAAC;QAC5C,MAAM,KAAK,CAAC,iEAAiE;YACzE,yBAAyB,UAAU,KAAK,SAAS,YAAY,MAAM,KAAK,MAAM,IAAI,CAAC,CAAC;IAC5F,CAAC;IAED,qEAAqE;IACrE,qEAAqE;IACrE,OAAO;QACH,uBAAuB;QACvB,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,MAAM,EAAE,GAAG,CAAC,UAAU,GAAG,MAAM,GAAG,CAAC,CAAC,CAAC;QACrD,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,MAAM,EAAE,GAAG,CAAC,SAAS,GAAG,MAAM,GAAG,CAAC,CAAC,CAAC;QACpD,CAAC,CAAC,yDAAyD;KAC9D,CAAA;AACL,CAAC;AAGD;;;;;;;GAOG;AACH,MAAM,UAAU,yBAAyB,CAAC,KAAkB,EAAE,KAAa;IACvE,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,KAAK,CAAC,CAAC;IAErC,MAAM,CAAC,MAAM,EAAE,KAAK,CAAC,GAAG,KAAK,CAAC,KAAK,CAAC;IAEpC,OAAO;QACH,CAAC,IAAI,CAAC,IAAI,CAAC,MAAM,GAAG,SAAS,CAAC,CAAC,GAAG,SAAS,GAAG,MAAM;QACpD,CAAC,IAAI,CAAC,IAAI,CAAC,KAAK,GAAG,SAAS,CAAC,CAAC,GAAG,SAAS,GAAG,KAAK;KACrD,CAAA;AACL,CAAC"}

package/dist/src/utils.test.d.ts

@@ -0,0 +1,2 @@
+export {};
+//# sourceMappingURL=utils.test.d.ts.map

package/dist/src/utils.test.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"utils.test.d.ts","sourceRoot":"","sources":["../../src/utils.test.ts"],"names":[],"mappings":""}

package/{src/utils.test.ts → dist/src/utils.test.js}

@@ -1,90 +1,63 @@
 import * as tf from "@tensorflow/tfjs";
-import { getScaleShape, getRandomCropStart, generateCausalAttentionMask } from "@/utils";
-
+import { getScaleShape, getRandomCropStart } from "@/utils";
+import { causal } from "@/masks";
 // avoid TFJS node message during Jest testing
 tf.env().set('IS_NODE', false);
-
-
 describe("test custom TFJS utility functions", () => {
-
     test("crop an image using the same shape, results in same shape", async () => {
         // cropping an image of the same shape
-        const img_size = [133, 84] as [number, number];
-        const target_size = [133, 84] as [number, number];
-
+        const img_size = [133, 84];
+        const target_size = [133, 84];
         expect(getRandomCropStart(img_size, target_size)).toEqual([0, 0, 0]);
     });
-
-
     it("should throw when crop is larger than image", async () => {
         expect(() => getRandomCropStart([128, 128], [1000, 2000])).toThrow();
-    })
-
-
+    });
     test("cropped image shape", async () => {
         // cropping from wide to tall image
         for (let i = 0; i < 100; i++) {
-            const img_size = [4923, 832] as [number, number];
-            const target_size = [333, 739] as [number, number];
-
-            const [crop_start_h, crop_start_w, channels] = getRandomCropStart(img_size, target_size)
-
+            const img_size = [4923, 832];
+            const target_size = [333, 739];
+            const [crop_start_h, crop_start_w, channels] = getRandomCropStart(img_size, target_size);
             expect(crop_start_h).toBeLessThanOrEqual(img_size[0] - target_size[0]);
             expect(crop_start_w).toBeLessThanOrEqual(img_size[1] - target_size[1]);
         }
-
         // cropping from tall to wide image
         for (let i = 0; i < 100; i++) {
-            const img_size = [381, 999] as [number, number];
-            const target_size = [300, 157] as [number, number];
-
-            const [crop_start_h, crop_start_w, channels] = getRandomCropStart(img_size, target_size)
-
+            const img_size = [381, 999];
+            const target_size = [300, 157];
+            const [crop_start_h, crop_start_w, channels] = getRandomCropStart(img_size, target_size);
             expect(crop_start_h).toBeLessThanOrEqual(img_size[0] - target_size[0]);
             expect(crop_start_w).toBeLessThanOrEqual(img_size[1] - target_size[1]);
         }
     });
-
-
     test("scale 1:1, results in the same shape", async () => {
-        const scale = getScaleShape([256, 256], [256, 256])
+        const scale = getScaleShape([256, 256], [256, 256]);
         expect(scale).toEqual([256, 256]);
     });
-
-
     test("scaled image shape", async () => {
         // scaling squares result in squares
-        const scale1 = getScaleShape([256, 256], [128, 128])
+        const scale1 = getScaleShape([256, 256], [128, 128]);
         expect(scale1).toEqual([128, 128]);
-
-        const scale2 = getScaleShape([128, 128], [256, 256])
+        const scale2 = getScaleShape([128, 128], [256, 256]);
         expect(scale2).toEqual([256, 256]);
-
-        const scale3 = getScaleShape([123, 123], [321, 321])
+        const scale3 = getScaleShape([123, 123], [321, 321]);
         expect(scale3).toEqual([321, 321]);
-
-        const scale4 = getScaleShape([321, 321], [123, 123])
+        const scale4 = getScaleShape([321, 321], [123, 123]);
         expect(scale4).toEqual([123, 123]);
-
         // scaling rectangles result in rectangles
-        const scale5 = getScaleShape([640, 480], [1280, 960])
+        const scale5 = getScaleShape([640, 480], [1280, 960]);
         expect(scale5).toEqual([1280, 960]);
-
-        const scale6 = getScaleShape([480, 640], [960, 1280])
+        const scale6 = getScaleShape([480, 640], [960, 1280]);
         expect(scale6).toEqual([960, 1280]);
-
-        const [scale7_h, scale7_w] = getScaleShape([777, 555], [555, 333])
+        const [scale7_h, scale7_w] = getScaleShape([777, 555], [555, 333]);
         expect(scale7_h).toBeGreaterThan(scale7_w);
-
-        const [scale8_h, scale8_w] = getScaleShape([555, 777], [333, 555])
+        const [scale8_h, scale8_w] = getScaleShape([555, 777], [333, 555]);
         expect(scale8_h).toBeLessThan(scale8_w);
     });
-
-
     test("causal attention map", async () => {
         const seq_len = 4;
-        const causal_mask = generateCausalAttentionMask(seq_len, seq_len);
-
+        const causal_mask = causal(seq_len, seq_len);
         const _ = -1e7;
         const expected_mask = tf.tensor([
             [0, _, _, _],
@@ -92,10 +65,9 @@ describe("test custom TFJS utility functions", () => {
             [0, 0, 0, _],
             [0, 0, 0, 0]
         ]);
-
         // this might fail due to precision issues on the masked positions,
         // in which case use less <= to 6 or 12 (number of masked positions x2)
         expect((await causal_mask.sub(expected_mask).sum().data())[0]).toEqual(0);
     });
-
 });
+//# sourceMappingURL=utils.test.js.map

package/dist/src/utils.test.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"utils.test.js","sourceRoot":"","sources":["../../src/utils.test.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,kBAAkB,CAAC;AACvC,OAAO,EAAE,aAAa,EAAE,kBAAkB,EAAE,MAAM,SAAS,CAAC;AAC5D,OAAO,EAAE,MAAM,EAAE,MAAM,SAAS,CAAC;AAEjC,8CAA8C;AAC9C,EAAE,CAAC,GAAG,EAAE,CAAC,GAAG,CAAC,SAAS,EAAE,KAAK,CAAC,CAAC;AAG/B,QAAQ,CAAC,oCAAoC,EAAE,GAAG,EAAE;IAEhD,IAAI,CAAC,2DAA2D,EAAE,KAAK,IAAI,EAAE;QACzE,sCAAsC;QACtC,MAAM,QAAQ,GAAG,CAAC,GAAG,EAAE,EAAE,CAAqB,CAAC;QAC/C,MAAM,WAAW,GAAG,CAAC,GAAG,EAAE,EAAE,CAAqB,CAAC;QAElD,MAAM,CAAC,kBAAkB,CAAC,QAAQ,EAAE,WAAW,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;IACzE,CAAC,CAAC,CAAC;IAGH,EAAE,CAAC,6CAA6C,EAAE,KAAK,IAAI,EAAE;QACzD,MAAM,CAAC,GAAG,EAAE,CAAC,kBAAkB,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,EAAE,CAAC,IAAI,EAAE,IAAI,CAAC,CAAC,CAAC,CAAC,OAAO,EAAE,CAAC;IACzE,CAAC,CAAC,CAAA;IAGF,IAAI,CAAC,qBAAqB,EAAE,KAAK,IAAI,EAAE;QACnC,mCAAmC;QACnC,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,GAAG,EAAE,CAAC,EAAE,EAAE,CAAC;YAC3B,MAAM,QAAQ,GAAG,CAAC,IAAI,EAAE,GAAG,CAAqB,CAAC;YACjD,MAAM,WAAW,GAAG,CAAC,GAAG,EAAE,GAAG,CAAqB,CAAC;YAEnD,MAAM,CAAC,YAAY,EAAE,YAAY,EAAE,QAAQ,CAAC,GAAG,kBAAkB,CAAC,QAAQ,EAAE,WAAW,CAAC,CAAA;YAExF,MAAM,CAAC,YAAY,CAAC,CAAC,mBAAmB,CAAC,QAAQ,CAAC,CAAC,CAAC,GAAG,WAAW,CAAC,CAAC,CAAC,CAAC,CAAC;YACvE,MAAM,CAAC,YAAY,CAAC,CAAC,mBAAmB,CAAC,QAAQ,CAAC,CAAC,CAAC,GAAG,WAAW,CAAC,CAAC,CAAC,CAAC,CAAC;QAC3E,CAAC;QAED,mCAAmC;QACnC,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,GAAG,EAAE,CAAC,EAAE,EAAE,CAAC;YAC3B,MAAM,QAAQ,GAAG,CAAC,GAAG,EAAE,GAAG,CAAqB,CAAC;YAChD,MAAM,WAAW,GAAG,CAAC,GAAG,EAAE,GAAG,CAAqB,CAAC;YAEnD,MAAM,CAAC,YAAY,EAAE,YAAY,EAAE,QAAQ,CAAC,GAAG,kBAAkB,CAAC,QAAQ,EAAE,WAAW,CAAC,CAAA;YAExF,MAAM,CAAC,YAAY,CAAC,CAAC,mBAAmB,CAAC,QAAQ,CAAC,CAAC,CAAC,GAAG,WAAW,CAAC,CAAC,CAAC,CAAC,CAAC;YACvE,MAAM,CAAC,YAAY,CAAC,CAAC,mBAAmB,CAAC,QAAQ,CAAC,CAAC,CAAC,GAAG,WAAW,CAAC,CAAC,CAAC,CAAC,CAAC;QAC3E,CAAC;IACL,CAAC,CAAC,CAAC;IAGH,IAAI,CAAC,sCAAsC,EAAE,KAAK,IAAI,EAAE;QACpD,MAAM,KAAK,GAAG,aAAa,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,EAAE,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAA;QACnD,MAAM,CAAC,KAAK,CAAC,CAAC,OAAO,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;IACtC,CAAC,CAAC,CAAC;IAGH,IAAI,CAAC,oBAAoB,EAAE,KAAK,IAAI,EAAE;QAClC,oCAAoC;QACpC,MAAM,MAAM,GAAG,aAAa,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,EAAE,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAA;QACpD,MAAM,CAAC,MAAM,CAAC,CAAC,OAAO,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;QAEnC,MAAM,MAAM,GAAG,aAAa,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,EAAE,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAA;QACpD,MAAM,CAAC,MAAM,CAAC,CAAC,OAAO,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;QAEnC,MAAM,MAAM,GAAG,aAAa,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,EAAE,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAA;QACpD,MAAM,CAAC,MAAM,CAAC,CAAC,OAAO,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;QAEnC,MAAM,MAAM,GAAG,aAAa,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,EAAE,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAA;QACpD,MAAM,CAAC,MAAM,CAAC,CAAC,OAAO,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;QAEnC,0CAA0C;QAC1C,MAAM,MAAM,GAAG,aAAa,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,EAAE,CAAC,IAAI,EAAE,GAAG,CAAC,CAAC,CAAA;QACrD,MAAM,CAAC,MAAM,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI,EAAE,GAAG,CAAC,CAAC,CAAC;QAEpC,MAAM,MAAM,GAAG,aAAa,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,EAAE,CAAC,GAAG,EAAE,IAAI,CAAC,CAAC,CAAA;QACrD,MAAM,CAAC,MAAM,CAAC,CAAC,OAAO,CAAC,CAAC,GAAG,EAAE,IAAI,CAAC,CAAC,CAAC;QAEpC,MAAM,CAAC,QAAQ,EAAE,QAAQ,CAAC,GAAG,aAAa,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,EAAE,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAA;QAClE,MAAM,CAAC,QAAQ,CAAC,CAAC,eAAe,CAAC,QAAQ,CAAC,CAAC;QAE3C,MAAM,CAAC,QAAQ,EAAE,QAAQ,CAAC,GAAG,aAAa,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,EAAE,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAA;QAClE,MAAM,CAAC,QAAQ,CAAC,CAAC,YAAY,CAAC,QAAQ,CAAC,CAAC;IAC5C,CAAC,CAAC,CAAC;IAGH,IAAI,CAAC,sBAAsB,EAAE,KAAK,IAAI,EAAE;QACpC,MAAM,OAAO,GAAG,CAAC,CAAC;QAClB,MAAM,WAAW,GAAG,MAAM,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC;QAE7C,MAAM,CAAC,GAAG,CAAC,GAAG,CAAC;QACf,MAAM,aAAa,GAAG,EAAE,CAAC,MAAM,CAAC;YAC5B,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC;YACZ,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC;YACZ,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC;YACZ,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC;SACf,CAAC,CAAC;QAEH,mEAAmE;QACnE,uEAAuE;QACvE,MAAM,CAAC,CAAC,MAAM,WAAW,CAAC,GAAG,CAAC,aAAa,CAAC,CAAC,GAAG,EAAE,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC;IAC9E,CAAC,CAAC,CAAC;AAEP,CAAC,CAAC,CAAC"}

package/dist/tfjs_types.d.ts

@@ -0,0 +1,10 @@
+import type { Tensor } from "@tensorflow/tfjs";
+export declare abstract class LazyIterator<T> {
+    abstract next(): Promise<IteratorResult<T>>;
+}
+export declare abstract class Dataset<T> {
+    abstract iterator(): Promise<LazyIterator<T>>;
+    size: number;
+}
+export type LossOrMetricFn = (yTrue: Tensor, yPred: Tensor) => Tensor;
+//# sourceMappingURL=tfjs_types.d.ts.map

package/dist/tfjs_types.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"tfjs_types.d.ts","sourceRoot":"","sources":["../src/tfjs_types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,EAAE,MAAM,kBAAkB,CAAC;AAG/C,MAAM,CAAC,OAAO,CAAC,QAAQ,OAAO,YAAY,CAAC,CAAC;IACxC,QAAQ,CAAC,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC;CAC9C;AAGD,MAAM,CAAC,OAAO,CAAC,QAAQ,OAAO,OAAO,CAAC,CAAC;IACnC,QAAQ,CAAC,QAAQ,IAAI,OAAO,CAAC,YAAY,CAAC,CAAC,CAAC,CAAC;IAC7C,IAAI,EAAE,MAAM,CAAC;CAChB;AAGD,MAAM,MAAM,cAAc,GAAG,CAAC,KAAK,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,KAAK,MAAM,CAAC"}

package/dist/tfjs_types.js

@@ -0,0 +1,2 @@
+export {};
+//# sourceMappingURL=tfjs_types.js.map

package/dist/tfjs_types.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"tfjs_types.js","sourceRoot":"","sources":["../src/tfjs_types.ts"],"names":[],"mappings":""}

package/dist/utils.d.ts

@@ -0,0 +1,28 @@
+import * as tf from "@tensorflow/tfjs";
+/**
+ * Calculate the desired scaled image's height and width. The shortest edge will
+ * be scaled to match its corresponding target shape's edge. The longer
+ * edge might end up larger than intended.
+ *
+ * @param image_shape the `[height, width]` of the image
+ * @param target_shape the intended `[height, width]` of the final scaled image
+ */
+export declare function getScaleShape(image_shape: tf.Shape, target_shape: [number, number]): [scaled_height: number, scaled_width: number];
+/**
+ * Calculate the starting point for a crop (slice) operation
+ * on an image tensor with the shape `[height, width, channels]`.
+ *
+ * @param image_shape the `[height, width]` of the image
+ * @param target_shape the intended `[height, width]` of the final cropped image
+ */
+export declare function getRandomCropStart(image_shape: [height: number, width: number], target_shape: [height: number, width: number]): [number, number, number];
+/**
+ * Calculate the height and width padding such that the image is
+ * divisible by 2^depth.
+ *
+ * In U-Net image segmentation, the contraction and concatenate
+ * operations requires  the input image's height and width
+ * dimensions to be divisible by 2^depth.
+ */
+export declare function getPaddingForSegmentation(image: tf.Tensor3D, depth: number): [height: number, width: number];
+//# sourceMappingURL=utils.d.ts.map

package/dist/utils.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"utils.d.ts","sourceRoot":"","sources":["../src/utils.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,kBAAkB,CAAC;AAGvC;;;;;;;GAOG;AACH,wBAAgB,aAAa,CAAC,WAAW,EAAE,EAAE,CAAC,KAAK,EAAE,YAAY,EAAE,CAAC,MAAM,EAAE,MAAM,CAAC,GAAG,CAAC,aAAa,EAAE,MAAM,EAAE,YAAY,EAAE,MAAM,CAAC,CAelI;AAGD;;;;;;GAMG;AACH,wBAAgB,kBAAkB,CAC9B,WAAW,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,CAAC,EAC5C,YAAY,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,CAAC,GAC9C,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,CAiB1B;AAGD;;;;;;;GAOG;AACH,wBAAgB,yBAAyB,CAAC,KAAK,EAAE,EAAE,CAAC,QAAQ,EAAE,KAAK,EAAE,MAAM,GAAG,CAAC,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,CAAC,CAS5G"}

package/dist/utils.js

@@ -0,0 +1,63 @@
+/**
+ * Calculate the desired scaled image's height and width. The shortest edge will
+ * be scaled to match its corresponding target shape's edge. The longer
+ * edge might end up larger than intended.
+ *
+ * @param image_shape the `[height, width]` of the image
+ * @param target_shape the intended `[height, width]` of the final scaled image
+ */
+export function getScaleShape(image_shape, target_shape) {
+    const [img_height, img_width] = image_shape;
+    const [target_height, target_width] = target_shape;
+    // scale based on whichever target_edge / original_edge is largest,
+    // we need the following to be true (1)
+    // height * scale >= target_height
+    // width * scale >= target_height
+    // rearranging to get an equivalent requirement (2)
+    // scale >= target_height / height
+    // scale >= target_width / width
+    // by picking the scale value that's largest of the two, we satisfy (2), and therefore (1)
+    // it may be more intuitive to think of scale as scale_h and scale_w
+    const scale_factor = Math.max(target_height / img_height, target_width / img_width);
+    return [Math.round(img_height * scale_factor), Math.round(img_width * scale_factor)];
+}
+/**
+ * Calculate the starting point for a crop (slice) operation
+ * on an image tensor with the shape `[height, width, channels]`.
+ *
+ * @param image_shape the `[height, width]` of the image
+ * @param target_shape the intended `[height, width]` of the final cropped image
+ */
+export function getRandomCropStart(image_shape, target_shape) {
+    const [img_height, img_width] = image_shape;
+    const [crop_x, crop_y] = target_shape;
+    if (img_height < crop_x || img_width < crop_y) {
+        throw Error(`getRandomCropShape: cannot crop with a size that's bigger than,` +
+            ` the image. Original [${img_height}, ${img_width}], crop [${crop_x}, ${crop_y}].`);
+    }
+    // there's a +1 because Math.random()'s range is [0, 1), excluding 1,
+    // hence +1 to ensure the full range of possible crop starting points
+    return [
+        // TODO: revisit the +1
+        Math.floor(Math.random() * (img_height - crop_x + 1)),
+        Math.floor(Math.random() * (img_width - crop_y + 1)),
+        0 // not cropping channels, so it starts at the first index
+    ];
+}
+/**
+ * Calculate the height and width padding such that the image is
+ * divisible by 2^depth.
+ *
+ * In U-Net image segmentation, the contraction and concatenate
+ * operations requires  the input image's height and width
+ * dimensions to be divisible by 2^depth.
+ */
+export function getPaddingForSegmentation(image, depth) {
+    const divisible = Math.pow(2, depth);
+    const [height, width] = image.shape;
+    return [
+        (Math.ceil(height / divisible)) * divisible - height,
+        (Math.ceil(width / divisible)) * divisible - width,
+    ];
+}
+//# sourceMappingURL=utils.js.map

package/dist/utils.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"utils.js","sourceRoot":"","sources":["../src/utils.ts"],"names":[],"mappings":"AAGA;;;;;;;GAOG;AACH,MAAM,UAAU,aAAa,CAAC,WAAqB,EAAE,YAA8B;IAC/E,MAAM,CAAC,UAAU,EAAE,SAAS,CAAC,GAAG,WAAuC,CAAC;IACxE,MAAM,CAAC,aAAa,EAAE,YAAY,CAAC,GAAG,YAAY,CAAC;IAEnD,mEAAmE;IACnE,uCAAuC;IACvC,kCAAkC;IAClC,iCAAiC;IACjC,mDAAmD;IACnD,kCAAkC;IAClC,gCAAgC;IAChC,0FAA0F;IAC1F,oEAAoE;IACpE,MAAM,YAAY,GAAG,IAAI,CAAC,GAAG,CAAC,aAAa,GAAG,UAAU,EAAE,YAAY,GAAG,SAAS,CAAC,CAAC;IACpF,OAAO,CAAC,IAAI,CAAC,KAAK,CAAC,UAAU,GAAG,YAAY,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,SAAS,GAAG,YAAY,CAAC,CAAC,CAAC;AACzF,CAAC;AAGD;;;;;;GAMG;AACH,MAAM,UAAU,kBAAkB,CAC9B,WAA4C,EAC5C,YAA6C;IAE7C,MAAM,CAAC,UAAU,EAAE,SAAS,CAAC,GAAG,WAAW,CAAC;IAC5C,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,GAAG,YAAY,CAAC;IAEtC,IAAI,UAAU,GAAG,MAAM,IAAI,SAAS,GAAG,MAAM,EAAE,CAAC;QAC5C,MAAM,KAAK,CAAC,iEAAiE;YACzE,yBAAyB,UAAU,KAAK,SAAS,YAAY,MAAM,KAAK,MAAM,IAAI,CAAC,CAAC;IAC5F,CAAC;IAED,qEAAqE;IACrE,qEAAqE;IACrE,OAAO;QACH,uBAAuB;QACvB,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,MAAM,EAAE,GAAG,CAAC,UAAU,GAAG,MAAM,GAAG,CAAC,CAAC,CAAC;QACrD,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,MAAM,EAAE,GAAG,CAAC,SAAS,GAAG,MAAM,GAAG,CAAC,CAAC,CAAC;QACpD,CAAC,CAAC,yDAAyD;KAC9D,CAAA;AACL,CAAC;AAGD;;;;;;;GAOG;AACH,MAAM,UAAU,yBAAyB,CAAC,KAAkB,EAAE,KAAa;IACvE,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,KAAK,CAAC,CAAC;IAErC,MAAM,CAAC,MAAM,EAAE,KAAK,CAAC,GAAG,KAAK,CAAC,KAAK,CAAC;IAEpC,OAAO;QACH,CAAC,IAAI,CAAC,IAAI,CAAC,MAAM,GAAG,SAAS,CAAC,CAAC,GAAG,SAAS,GAAG,MAAM;QACpD,CAAC,IAAI,CAAC,IAAI,CAAC,KAAK,GAAG,SAAS,CAAC,CAAC,GAAG,SAAS,GAAG,KAAK;KACrD,CAAA;AACL,CAAC"}

package/dist/utils.test.d.ts

@@ -0,0 +1,2 @@
+export {};
+//# sourceMappingURL=utils.test.d.ts.map

package/dist/utils.test.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"utils.test.d.ts","sourceRoot":"","sources":["../src/utils.test.ts"],"names":[],"mappings":""}