@zh-keyboard/recognizer 0.2.0 → 0.2.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +4 -1
- package/dist/index.d.mts +1 -1
- package/dist/index.js +2 -2
- package/dist/index.js.map +1 -1
- package/dist/index.mjs +2 -2
- package/dist/index.mjs.map +1 -1
- package/package.json +3 -3
package/README.md
CHANGED
|
@@ -34,12 +34,14 @@ pnpm add @zh-keyboard/recognizer
|
|
|
34
34
|
|
|
35
35
|
```typescript
|
|
36
36
|
import { ZhkRecognizer } from '@zh-keyboard/recognizer'
|
|
37
|
+
// 如果需要使用WebGL后端，需要手动引入
|
|
38
|
+
import '@tensorflow/tfjs-backend-webgl'
|
|
37
39
|
|
|
38
40
|
// 创建识别器实例
|
|
39
41
|
const recognizer = new ZhkRecognizer({
|
|
40
42
|
modelPath: '/models/handwrite/model.json', // TensorFlow.js模型路径
|
|
41
43
|
dictPath: '/models/dict.txt', // 汉字字典路径
|
|
42
|
-
backend: 'webgl' // 可选：'webgl'或'cpu'，默认为'
|
|
44
|
+
backend: 'webgl' // 可选：'webgl'或'cpu'，默认为'cpu'
|
|
43
45
|
})
|
|
44
46
|
|
|
45
47
|
// 初始化识别器（加载模型和字典）
|
|
@@ -68,5 +70,6 @@ await recognizer.close()
|
|
|
68
70
|
## 性能优化
|
|
69
71
|
|
|
70
72
|
- 在支持WebGL的设备上使用'webgl'后端可以显著提高识别速度
|
|
73
|
+
- **注意：使用WebGL后端时，需要手动引入 `import '@tensorflow/tfjs-backend-webgl'`**
|
|
71
74
|
- 初始化时会自动进行模型预热，减少首次识别的延迟
|
|
72
75
|
- 如果在移动设备上使用，建议使用较小的模型以提高响应速度
|
package/dist/index.d.mts
CHANGED
package/dist/index.js
CHANGED
|
@@ -24,7 +24,7 @@ var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__ge
|
|
|
24
24
|
//#endregion
|
|
25
25
|
const __tensorflow_tfjs_converter = __toESM(require("@tensorflow/tfjs-converter"));
|
|
26
26
|
const __tensorflow_tfjs_core = __toESM(require("@tensorflow/tfjs-core"));
|
|
27
|
-
require("@tensorflow/tfjs-backend-
|
|
27
|
+
require("@tensorflow/tfjs-backend-cpu");
|
|
28
28
|
|
|
29
29
|
//#region src/index.ts
|
|
30
30
|
var ZhkRecognizer = class {
|
|
@@ -38,7 +38,7 @@ var ZhkRecognizer = class {
|
|
|
38
38
|
constructor(options) {
|
|
39
39
|
this.modelPath = options.modelPath;
|
|
40
40
|
this.dictPath = options.dictPath;
|
|
41
|
-
this.backend = options.backend || "
|
|
41
|
+
this.backend = options.backend || "cpu";
|
|
42
42
|
this.canvas = document.createElement("canvas");
|
|
43
43
|
this.canvas.width = this.canvas.height = 64;
|
|
44
44
|
this.ctx = this.canvas.getContext("2d", { willReadFrequently: true });
|
package/dist/index.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.js","names":["options: RecognizerOptions","options?: RecognizerInitOptions","strokeData: number[]","last: { x: number, y: number, isEnd: boolean } | null"],"sources":["../src/index.ts"],"sourcesContent":["import type { GraphModel, Tensor } from '@tensorflow/tfjs'\r\nimport type { HandwritingRecognizer, RecognizerInitOptions } from '@zh-keyboard/core'\r\nimport { loadGraphModel } from '@tensorflow/tfjs-converter'\r\nimport * as tf from '@tensorflow/tfjs-core'\r\nimport '@tensorflow/tfjs-backend-
|
|
1
|
+
{"version":3,"file":"index.js","names":["options: RecognizerOptions","options?: RecognizerInitOptions","strokeData: number[]","last: { x: number, y: number, isEnd: boolean } | null"],"sources":["../src/index.ts"],"sourcesContent":["import type { GraphModel, Tensor } from '@tensorflow/tfjs'\r\nimport type { HandwritingRecognizer, RecognizerInitOptions } from '@zh-keyboard/core'\r\nimport { loadGraphModel } from '@tensorflow/tfjs-converter'\r\nimport * as tf from '@tensorflow/tfjs-core'\r\nimport '@tensorflow/tfjs-backend-cpu'\r\n\r\nexport interface RecognizerOptions {\r\n /**\r\n * 模型路径\r\n */\r\n modelPath: string\r\n /**\r\n * 字典路径\r\n */\r\n dictPath: string\r\n /**\r\n * 后端类型\r\n */\r\n backend?: 'webgl' | 'cpu'\r\n}\r\n\r\nexport class ZhkRecognizer implements HandwritingRecognizer {\r\n private model?: GraphModel\r\n private dict: string[] = []\r\n private canvas: HTMLCanvasElement\r\n private ctx: CanvasRenderingContext2D\r\n private modelPath: string\r\n private dictPath: string\r\n private backend: 'webgl' | 'cpu'\r\n\r\n constructor(options: RecognizerOptions) {\r\n this.modelPath = options.modelPath\r\n this.dictPath = options.dictPath\r\n this.backend = options.backend || 'cpu'\r\n this.canvas = document.createElement('canvas')\r\n this.canvas.width = this.canvas.height = 64\r\n this.ctx = this.canvas.getContext('2d', { willReadFrequently: true })!\r\n }\r\n\r\n async initialize(options?: RecognizerInitOptions) {\r\n const text = await fetch(this.dictPath).then(r => r.text())\r\n this.dict = text.split('\\n')\r\n this.model = await loadGraphModel(this.modelPath, {\r\n streamWeights: true,\r\n onProgress: options?.onProgress,\r\n })\r\n // 如果后端为webgl，则需要进行预热\r\n if (this.backend === 'webgl') {\r\n await tf.setBackend('webgl')\r\n await tf.ready()\r\n await this.recognize([10, 10, 0, 20, 20, 1])\r\n } else {\r\n await tf.setBackend('cpu')\r\n }\r\n return true\r\n }\r\n\r\n async recognize(strokeData: number[]): Promise<string[]> {\r\n if (!this.model) {\r\n throw new Error('Model not initialized')\r\n }\r\n const { canvas, ctx, model, dict } = this\r\n ctx.fillStyle = 'white'\r\n ctx.fillRect(0, 0, canvas.width, canvas.height)\r\n\r\n const n = strokeData.length / 3\r\n const strokes = Array.from({ length: n }, (_, i) => ({\r\n x: strokeData[3 * i],\r\n y: strokeData[3 * i + 1],\r\n isEnd: strokeData[3 * i + 2] === 1,\r\n }))\r\n\r\n let minX = Infinity\r\n let minY = Infinity\r\n let maxX = -Infinity\r\n let maxY = -Infinity\r\n for (const { x, y } of strokes) {\r\n if (x < minX) {\r\n minX = x\r\n }\r\n if (x > maxX) {\r\n maxX = x\r\n }\r\n if (y < minY) {\r\n minY = y\r\n }\r\n if (y > maxY) {\r\n maxY = y\r\n }\r\n }\r\n\r\n const w = maxX - minX || 1\r\n const h = maxY - minY || 1\r\n const cx = (minX + maxX) / 2\r\n const cy = (minY + maxY) / 2\r\n const scale = Math.min(canvas.width * 0.9 / w, canvas.height * 0.9 / h)\r\n\r\n ctx.strokeStyle = 'black'\r\n ctx.lineWidth = 2\r\n ctx.lineCap = 'round'\r\n ctx.lineJoin = 'round'\r\n\r\n let last: { x: number, y: number, isEnd: boolean } | null = null\r\n for (const s of strokes) {\r\n const x = canvas.width / 2 + (s.x - cx) * scale\r\n const y = canvas.height / 2 + (s.y - cy) * scale\r\n if (last && !last.isEnd) {\r\n ctx.beginPath()\r\n ctx.moveTo(canvas.width / 2 + (last.x - cx) * scale, canvas.height / 2 + (last.y - cy) * scale)\r\n ctx.lineTo(x, y)\r\n ctx.stroke()\r\n } else {\r\n ctx.beginPath()\r\n ctx.moveTo(x, y)\r\n }\r\n last = s\r\n }\r\n\r\n return tf.tidy(() => {\r\n const image = tf.browser.fromPixels(canvas, 3)\r\n const floatImage = tf.cast(image, 'float32')\r\n const normalizedImage = tf.div(floatImage, 255)\r\n const batchedImage = tf.expandDims(normalizedImage, 0)\r\n\r\n const probs = (model!.predict(batchedImage) as Tensor).dataSync()\r\n const idxs = Array.from(probs.keys()).sort((a, b) => probs[b] - probs[a]).slice(0, 10)\r\n\r\n return idxs.map(i => (i < dict.length ? dict[i] : '')).filter(Boolean)\r\n })\r\n }\r\n\r\n async close() {\r\n this.model?.dispose()\r\n this.model = undefined\r\n }\r\n}\r\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAqBA,IAAa,gBAAb,MAA4D;CAC1D,AAAQ;CACR,AAAQ,OAAiB,CAAE;CAC3B,AAAQ;CACR,AAAQ;CACR,AAAQ;CACR,AAAQ;CACR,AAAQ;CAER,YAAYA,SAA4B;AACtC,OAAK,YAAY,QAAQ;AACzB,OAAK,WAAW,QAAQ;AACxB,OAAK,UAAU,QAAQ,WAAW;AAClC,OAAK,SAAS,SAAS,cAAc,SAAS;AAC9C,OAAK,OAAO,QAAQ,KAAK,OAAO,SAAS;AACzC,OAAK,MAAM,KAAK,OAAO,WAAW,MAAM,EAAE,oBAAoB,KAAM,EAAC;CACtE;CAED,MAAM,WAAWC,SAAiC;EAChD,MAAM,OAAO,MAAM,MAAM,KAAK,SAAS,CAAC,KAAK,OAAK,EAAE,MAAM,CAAC;AAC3D,OAAK,OAAO,KAAK,MAAM,KAAK;AAC5B,OAAK,QAAQ,MAAM,gDAAe,KAAK,WAAW;GAChD,eAAe;GACf,YAAY,SAAS;EACtB,EAAC;AAEF,MAAI,KAAK,YAAY,SAAS;AAC5B,SAAM,uBAAG,WAAW,QAAQ;AAC5B,SAAM,uBAAG,OAAO;AAChB,SAAM,KAAK,UAAU;IAAC;IAAI;IAAI;IAAG;IAAI;IAAI;GAAE,EAAC;EAC7C,MACC,OAAM,uBAAG,WAAW,MAAM;AAE5B,SAAO;CACR;CAED,MAAM,UAAUC,YAAyC;AACvD,OAAK,KAAK,MACR,OAAM,IAAI,MAAM;EAElB,MAAM,EAAE,QAAQ,KAAK,OAAO,MAAM,GAAG;AACrC,MAAI,YAAY;AAChB,MAAI,SAAS,GAAG,GAAG,OAAO,OAAO,OAAO,OAAO;EAE/C,MAAM,IAAI,WAAW,SAAS;EAC9B,MAAM,UAAU,MAAM,KAAK,EAAE,QAAQ,EAAG,GAAE,CAAC,GAAG,OAAO;GACnD,GAAG,WAAW,IAAI;GAClB,GAAG,WAAW,IAAI,IAAI;GACtB,OAAO,WAAW,IAAI,IAAI,OAAO;EAClC,GAAE;EAEH,IAAI,OAAO;EACX,IAAI,OAAO;EACX,IAAI,OAAO;EACX,IAAI,OAAO;AACX,OAAK,MAAM,EAAE,GAAG,GAAG,IAAI,SAAS;AAC9B,OAAI,IAAI,KACN,QAAO;AAET,OAAI,IAAI,KACN,QAAO;AAET,OAAI,IAAI,KACN,QAAO;AAET,OAAI,IAAI,KACN,QAAO;EAEV;EAED,MAAM,IAAI,OAAO,QAAQ;EACzB,MAAM,IAAI,OAAO,QAAQ;EACzB,MAAM,MAAM,OAAO,QAAQ;EAC3B,MAAM,MAAM,OAAO,QAAQ;EAC3B,MAAM,QAAQ,KAAK,IAAI,OAAO,QAAQ,KAAM,GAAG,OAAO,SAAS,KAAM,EAAE;AAEvE,MAAI,cAAc;AAClB,MAAI,YAAY;AAChB,MAAI,UAAU;AACd,MAAI,WAAW;EAEf,IAAIC,OAAwD;AAC5D,OAAK,MAAM,KAAK,SAAS;GACvB,MAAM,IAAI,OAAO,QAAQ,KAAK,EAAE,IAAI,MAAM;GAC1C,MAAM,IAAI,OAAO,SAAS,KAAK,EAAE,IAAI,MAAM;AAC3C,OAAI,SAAS,KAAK,OAAO;AACvB,QAAI,WAAW;AACf,QAAI,OAAO,OAAO,QAAQ,KAAK,KAAK,IAAI,MAAM,OAAO,OAAO,SAAS,KAAK,KAAK,IAAI,MAAM,MAAM;AAC/F,QAAI,OAAO,GAAG,EAAE;AAChB,QAAI,QAAQ;GACb,OAAM;AACL,QAAI,WAAW;AACf,QAAI,OAAO,GAAG,EAAE;GACjB;AACD,UAAO;EACR;AAED,SAAO,uBAAG,KAAK,MAAM;GACnB,MAAM,QAAQ,uBAAG,QAAQ,WAAW,QAAQ,EAAE;GAC9C,MAAM,aAAa,uBAAG,KAAK,OAAO,UAAU;GAC5C,MAAM,kBAAkB,uBAAG,IAAI,YAAY,IAAI;GAC/C,MAAM,eAAe,uBAAG,WAAW,iBAAiB,EAAE;GAEtD,MAAM,QAAQ,AAAC,MAAO,QAAQ,aAAa,CAAY,UAAU;GACjE,MAAM,OAAO,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC,KAAK,CAAC,GAAG,MAAM,MAAM,KAAK,MAAM,GAAG,CAAC,MAAM,GAAG,GAAG;AAEtF,UAAO,KAAK,IAAI,OAAM,IAAI,KAAK,SAAS,KAAK,KAAK,GAAI,CAAC,OAAO,QAAQ;EACvE,EAAC;CACH;CAED,MAAM,QAAQ;AACZ,OAAK,OAAO,SAAS;AACrB,OAAK;CACN;AACF"}
|
package/dist/index.mjs
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
import { loadGraphModel } from "@tensorflow/tfjs-converter";
|
|
2
2
|
import * as tf from "@tensorflow/tfjs-core";
|
|
3
|
-
import "@tensorflow/tfjs-backend-
|
|
3
|
+
import "@tensorflow/tfjs-backend-cpu";
|
|
4
4
|
|
|
5
5
|
//#region src/index.ts
|
|
6
6
|
var ZhkRecognizer = class {
|
|
@@ -14,7 +14,7 @@ var ZhkRecognizer = class {
|
|
|
14
14
|
constructor(options) {
|
|
15
15
|
this.modelPath = options.modelPath;
|
|
16
16
|
this.dictPath = options.dictPath;
|
|
17
|
-
this.backend = options.backend || "
|
|
17
|
+
this.backend = options.backend || "cpu";
|
|
18
18
|
this.canvas = document.createElement("canvas");
|
|
19
19
|
this.canvas.width = this.canvas.height = 64;
|
|
20
20
|
this.ctx = this.canvas.getContext("2d", { willReadFrequently: true });
|
package/dist/index.mjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.mjs","names":["options: RecognizerOptions","options?: RecognizerInitOptions","strokeData: number[]","last: { x: number, y: number, isEnd: boolean } | null"],"sources":["../src/index.ts"],"sourcesContent":["import type { GraphModel, Tensor } from '@tensorflow/tfjs'\r\nimport type { HandwritingRecognizer, RecognizerInitOptions } from '@zh-keyboard/core'\r\nimport { loadGraphModel } from '@tensorflow/tfjs-converter'\r\nimport * as tf from '@tensorflow/tfjs-core'\r\nimport '@tensorflow/tfjs-backend-
|
|
1
|
+
{"version":3,"file":"index.mjs","names":["options: RecognizerOptions","options?: RecognizerInitOptions","strokeData: number[]","last: { x: number, y: number, isEnd: boolean } | null"],"sources":["../src/index.ts"],"sourcesContent":["import type { GraphModel, Tensor } from '@tensorflow/tfjs'\r\nimport type { HandwritingRecognizer, RecognizerInitOptions } from '@zh-keyboard/core'\r\nimport { loadGraphModel } from '@tensorflow/tfjs-converter'\r\nimport * as tf from '@tensorflow/tfjs-core'\r\nimport '@tensorflow/tfjs-backend-cpu'\r\n\r\nexport interface RecognizerOptions {\r\n /**\r\n * 模型路径\r\n */\r\n modelPath: string\r\n /**\r\n * 字典路径\r\n */\r\n dictPath: string\r\n /**\r\n * 后端类型\r\n */\r\n backend?: 'webgl' | 'cpu'\r\n}\r\n\r\nexport class ZhkRecognizer implements HandwritingRecognizer {\r\n private model?: GraphModel\r\n private dict: string[] = []\r\n private canvas: HTMLCanvasElement\r\n private ctx: CanvasRenderingContext2D\r\n private modelPath: string\r\n private dictPath: string\r\n private backend: 'webgl' | 'cpu'\r\n\r\n constructor(options: RecognizerOptions) {\r\n this.modelPath = options.modelPath\r\n this.dictPath = options.dictPath\r\n this.backend = options.backend || 'cpu'\r\n this.canvas = document.createElement('canvas')\r\n this.canvas.width = this.canvas.height = 64\r\n this.ctx = this.canvas.getContext('2d', { willReadFrequently: true })!\r\n }\r\n\r\n async initialize(options?: RecognizerInitOptions) {\r\n const text = await fetch(this.dictPath).then(r => r.text())\r\n this.dict = text.split('\\n')\r\n this.model = await loadGraphModel(this.modelPath, {\r\n streamWeights: true,\r\n onProgress: options?.onProgress,\r\n })\r\n // 如果后端为webgl，则需要进行预热\r\n if (this.backend === 'webgl') {\r\n await tf.setBackend('webgl')\r\n await tf.ready()\r\n await this.recognize([10, 10, 0, 20, 20, 1])\r\n } else {\r\n await tf.setBackend('cpu')\r\n }\r\n return true\r\n }\r\n\r\n async recognize(strokeData: number[]): Promise<string[]> {\r\n if (!this.model) {\r\n throw new Error('Model not initialized')\r\n }\r\n const { canvas, ctx, model, dict } = this\r\n ctx.fillStyle = 'white'\r\n ctx.fillRect(0, 0, canvas.width, canvas.height)\r\n\r\n const n = strokeData.length / 3\r\n const strokes = Array.from({ length: n }, (_, i) => ({\r\n x: strokeData[3 * i],\r\n y: strokeData[3 * i + 1],\r\n isEnd: strokeData[3 * i + 2] === 1,\r\n }))\r\n\r\n let minX = Infinity\r\n let minY = Infinity\r\n let maxX = -Infinity\r\n let maxY = -Infinity\r\n for (const { x, y } of strokes) {\r\n if (x < minX) {\r\n minX = x\r\n }\r\n if (x > maxX) {\r\n maxX = x\r\n }\r\n if (y < minY) {\r\n minY = y\r\n }\r\n if (y > maxY) {\r\n maxY = y\r\n }\r\n }\r\n\r\n const w = maxX - minX || 1\r\n const h = maxY - minY || 1\r\n const cx = (minX + maxX) / 2\r\n const cy = (minY + maxY) / 2\r\n const scale = Math.min(canvas.width * 0.9 / w, canvas.height * 0.9 / h)\r\n\r\n ctx.strokeStyle = 'black'\r\n ctx.lineWidth = 2\r\n ctx.lineCap = 'round'\r\n ctx.lineJoin = 'round'\r\n\r\n let last: { x: number, y: number, isEnd: boolean } | null = null\r\n for (const s of strokes) {\r\n const x = canvas.width / 2 + (s.x - cx) * scale\r\n const y = canvas.height / 2 + (s.y - cy) * scale\r\n if (last && !last.isEnd) {\r\n ctx.beginPath()\r\n ctx.moveTo(canvas.width / 2 + (last.x - cx) * scale, canvas.height / 2 + (last.y - cy) * scale)\r\n ctx.lineTo(x, y)\r\n ctx.stroke()\r\n } else {\r\n ctx.beginPath()\r\n ctx.moveTo(x, y)\r\n }\r\n last = s\r\n }\r\n\r\n return tf.tidy(() => {\r\n const image = tf.browser.fromPixels(canvas, 3)\r\n const floatImage = tf.cast(image, 'float32')\r\n const normalizedImage = tf.div(floatImage, 255)\r\n const batchedImage = tf.expandDims(normalizedImage, 0)\r\n\r\n const probs = (model!.predict(batchedImage) as Tensor).dataSync()\r\n const idxs = Array.from(probs.keys()).sort((a, b) => probs[b] - probs[a]).slice(0, 10)\r\n\r\n return idxs.map(i => (i < dict.length ? dict[i] : '')).filter(Boolean)\r\n })\r\n }\r\n\r\n async close() {\r\n this.model?.dispose()\r\n this.model = undefined\r\n }\r\n}\r\n"],"mappings":";;;;;AAqBA,IAAa,gBAAb,MAA4D;CAC1D,AAAQ;CACR,AAAQ,OAAiB,CAAE;CAC3B,AAAQ;CACR,AAAQ;CACR,AAAQ;CACR,AAAQ;CACR,AAAQ;CAER,YAAYA,SAA4B;AACtC,OAAK,YAAY,QAAQ;AACzB,OAAK,WAAW,QAAQ;AACxB,OAAK,UAAU,QAAQ,WAAW;AAClC,OAAK,SAAS,SAAS,cAAc,SAAS;AAC9C,OAAK,OAAO,QAAQ,KAAK,OAAO,SAAS;AACzC,OAAK,MAAM,KAAK,OAAO,WAAW,MAAM,EAAE,oBAAoB,KAAM,EAAC;CACtE;CAED,MAAM,WAAWC,SAAiC;EAChD,MAAM,OAAO,MAAM,MAAM,KAAK,SAAS,CAAC,KAAK,OAAK,EAAE,MAAM,CAAC;AAC3D,OAAK,OAAO,KAAK,MAAM,KAAK;AAC5B,OAAK,QAAQ,MAAM,eAAe,KAAK,WAAW;GAChD,eAAe;GACf,YAAY,SAAS;EACtB,EAAC;AAEF,MAAI,KAAK,YAAY,SAAS;AAC5B,SAAM,GAAG,WAAW,QAAQ;AAC5B,SAAM,GAAG,OAAO;AAChB,SAAM,KAAK,UAAU;IAAC;IAAI;IAAI;IAAG;IAAI;IAAI;GAAE,EAAC;EAC7C,MACC,OAAM,GAAG,WAAW,MAAM;AAE5B,SAAO;CACR;CAED,MAAM,UAAUC,YAAyC;AACvD,OAAK,KAAK,MACR,OAAM,IAAI,MAAM;EAElB,MAAM,EAAE,QAAQ,KAAK,OAAO,MAAM,GAAG;AACrC,MAAI,YAAY;AAChB,MAAI,SAAS,GAAG,GAAG,OAAO,OAAO,OAAO,OAAO;EAE/C,MAAM,IAAI,WAAW,SAAS;EAC9B,MAAM,UAAU,MAAM,KAAK,EAAE,QAAQ,EAAG,GAAE,CAAC,GAAG,OAAO;GACnD,GAAG,WAAW,IAAI;GAClB,GAAG,WAAW,IAAI,IAAI;GACtB,OAAO,WAAW,IAAI,IAAI,OAAO;EAClC,GAAE;EAEH,IAAI,OAAO;EACX,IAAI,OAAO;EACX,IAAI,OAAO;EACX,IAAI,OAAO;AACX,OAAK,MAAM,EAAE,GAAG,GAAG,IAAI,SAAS;AAC9B,OAAI,IAAI,KACN,QAAO;AAET,OAAI,IAAI,KACN,QAAO;AAET,OAAI,IAAI,KACN,QAAO;AAET,OAAI,IAAI,KACN,QAAO;EAEV;EAED,MAAM,IAAI,OAAO,QAAQ;EACzB,MAAM,IAAI,OAAO,QAAQ;EACzB,MAAM,MAAM,OAAO,QAAQ;EAC3B,MAAM,MAAM,OAAO,QAAQ;EAC3B,MAAM,QAAQ,KAAK,IAAI,OAAO,QAAQ,KAAM,GAAG,OAAO,SAAS,KAAM,EAAE;AAEvE,MAAI,cAAc;AAClB,MAAI,YAAY;AAChB,MAAI,UAAU;AACd,MAAI,WAAW;EAEf,IAAIC,OAAwD;AAC5D,OAAK,MAAM,KAAK,SAAS;GACvB,MAAM,IAAI,OAAO,QAAQ,KAAK,EAAE,IAAI,MAAM;GAC1C,MAAM,IAAI,OAAO,SAAS,KAAK,EAAE,IAAI,MAAM;AAC3C,OAAI,SAAS,KAAK,OAAO;AACvB,QAAI,WAAW;AACf,QAAI,OAAO,OAAO,QAAQ,KAAK,KAAK,IAAI,MAAM,OAAO,OAAO,SAAS,KAAK,KAAK,IAAI,MAAM,MAAM;AAC/F,QAAI,OAAO,GAAG,EAAE;AAChB,QAAI,QAAQ;GACb,OAAM;AACL,QAAI,WAAW;AACf,QAAI,OAAO,GAAG,EAAE;GACjB;AACD,UAAO;EACR;AAED,SAAO,GAAG,KAAK,MAAM;GACnB,MAAM,QAAQ,GAAG,QAAQ,WAAW,QAAQ,EAAE;GAC9C,MAAM,aAAa,GAAG,KAAK,OAAO,UAAU;GAC5C,MAAM,kBAAkB,GAAG,IAAI,YAAY,IAAI;GAC/C,MAAM,eAAe,GAAG,WAAW,iBAAiB,EAAE;GAEtD,MAAM,QAAQ,AAAC,MAAO,QAAQ,aAAa,CAAY,UAAU;GACjE,MAAM,OAAO,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC,KAAK,CAAC,GAAG,MAAM,MAAM,KAAK,MAAM,GAAG,CAAC,MAAM,GAAG,GAAG;AAEtF,UAAO,KAAK,IAAI,OAAM,IAAI,KAAK,SAAS,KAAK,KAAK,GAAI,CAAC,OAAO,QAAQ;EACvE,EAAC;CACH;CAED,MAAM,QAAQ;AACZ,OAAK,OAAO,SAAS;AACrB,OAAK;CACN;AACF"}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@zh-keyboard/recognizer",
|
|
3
|
-
"version": "0.2.
|
|
3
|
+
"version": "0.2.2",
|
|
4
4
|
"description": "中文手写汉字识别器组件",
|
|
5
5
|
"author": "dusionlike",
|
|
6
6
|
"license": "Apache 2.0",
|
|
@@ -36,10 +36,10 @@
|
|
|
36
36
|
"qrcode.jpg"
|
|
37
37
|
],
|
|
38
38
|
"dependencies": {
|
|
39
|
-
"@tensorflow/tfjs-backend-
|
|
39
|
+
"@tensorflow/tfjs-backend-cpu": "^4.22.0",
|
|
40
40
|
"@tensorflow/tfjs-converter": "^4.22.0",
|
|
41
41
|
"@tensorflow/tfjs-core": "^4.22.0",
|
|
42
|
-
"@zh-keyboard/core": "0.2.
|
|
42
|
+
"@zh-keyboard/core": "0.2.2"
|
|
43
43
|
},
|
|
44
44
|
"devDependencies": {
|
|
45
45
|
"tsdown": "^0.11.1",
|