npm - @sssxyd/face-liveness-detector - Versions diffs - 0.4.0-alpha.6 → 0.4.0-alpha.8 - Mend

@sssxyd/face-liveness-detector 0.4.0-alpha.6 → 0.4.0-alpha.8

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.

Files changed (12) hide show

package/README.en.md +844 -0
package/README.md +544 -393
package/dist/index.esm.js +37 -18
package/dist/index.esm.js.map +1 -1
package/dist/index.js +37 -18
package/dist/index.js.map +1 -1
package/dist/types/motion-liveness-detector.d.ts +8 -3
package/dist/types/motion-liveness-detector.d.ts.map +1 -1
package/dist/types/types.d.ts +82 -0
package/dist/types/types.d.ts.map +1 -1
package/package.json +1 -1
package/README.zh-Hans.md +0 -695

package/README.md CHANGED Viewed

@@ -1,445 +1,609 @@
-> **Languages:** [English](#) · [中文](./README.zh-Hans.md)
+<div align="center">
-# Face Detection Engine
+> **语言 / Languages:** [中文](#) · [English](./README.en.md)
-A pure frontend, real-time face liveness detection engine built on **[Human.js](https://github.com/vladmandic/human)** and **[OpenCV.js](https://github.com/TechStark/opencv-js)**. This TypeScript-based npm package provides real-time face detection, dual liveness verification (silent + action-based), automatic best frame selection, and anti-spoofing capabilities - all running 100% in the browser with zero backend dependency.
+# 人脸活体检测引擎
-## Features
+<p>
+  <strong>基于 TensorFlow + OpenCV 的纯前端实时人脸活体检测解决方案</strong>
+</p>
-- 💯 **Pure Frontend Implementation** - Zero backend dependency, all processing runs locally in the browser
-- 🔬 **Hybrid TensorFlow + OpenCV Solution** - Combines TensorFlow.js for AI detection with OpenCV.js for image processing
-- 🧠 **Dual Detection Modes** - Both silent liveness detection and action-based detection (blink, mouth open, nod) with automatic best frame selection
-- ⚡ **Pure JavaScript & Event-Driven** - 100% TypeScript, reactive event architecture, seamless integration with any frontend framework (Vue, React, Angular, Svelte, or vanilla JS)
-- 🎯 **Comprehensive Face Analysis** - Real-time anti-spoofing, quality assessment, frontality detection, and blur detection
+<p>
+  <img alt="TypeScript" src="https://img.shields.io/badge/TypeScript-5.0+-3178c6?logo=typescript">
+  <img alt="NPM Package" src="https://img.shields.io/npm/v/@sssxyd/face-liveness-detector?label=npm&color=cb3837">
+  <img alt="License" src="https://img.shields.io/badge/license-MIT-green">
+</p>
-## 🚀 Try Online Demo
+</div>
-**[👉 Live Demo: https://face.lowtechsoft.com/](https://face.lowtechsoft.com/)**
+---
-Scan the QR code with your phone to test the detection engine right now:
+## ✨ 功能特性
-[![Face Liveness Detection Demo QR Code](https://raw.githubusercontent.com/sssxyd/face-liveness-detector/main/demos/vue-demo/vue-demo.png)](https://face.lowtechsoft.com/)
-## Installation
+<table>
+  <tr>
+    <td>💯 <strong>纯前端实现</strong><br/>零后端依赖，所有处理在浏览器中本地运行</td>
+    <td>🔬 <strong>混合 AI 方案</strong><br/>TensorFlow + OpenCV 深度融合</td>
+  </tr>
+  <tr>
+    <td>🧠 <strong>双重活体验证</strong><br/>静默检测 + 动作识别（眨眼、张嘴、点头）</td>
+    <td>⚡ <strong>事件驱动架构</strong><br/>100% TypeScript，与任何框架无缝集成</td>
+  </tr>
+  <tr>
+    <td>🎯 <strong>全维度分析</strong><br/>质量、正对度、运动分数、屏幕检测</td>
+    <td>🛡️ <strong>多维反欺骗</strong><br/>照片、屏幕视频、莫尔纹、RGB发光检测</td>
+  </tr>
+</table>
+---
+## 🚀 在线演示
+<div align="center">
+**[👉 实时体验演示](https://face.lowtechsoft.com/) | 用手机扫码快速测试**
+[![人脸活体检测演示二维码](https://raw.githubusercontent.com/sssxyd/face-liveness-detector/main/demos/vue-demo/vue-demo.png)](https://face.lowtechsoft.com/)
+</div>
+---
+## 🧬 核心算法设计
+| 检测模块 | 技术方案 | 说明文档 |
+|---------|--------|--------|
+| **人脸识别** | Human.js BlazeFace + FaceMesh | 468个面部特征点 + 表情识别 |
+| **动作检测** | 多维度运动分析 | [运动检测算法](./docs/MOTION_DETECTION_ALGORITHM.md) - 光流、关键点方差、面部区域变化 |
+| **屏幕检测** | 三维特征融合 | [屏幕捕捉检测](./docs/SCREEN_CAPTURE_DETECTION_ALGORITHM.md) - 莫尔纹、RGB发光、色彩特征 |
+---
+## 📦 安装指南
+### 快速安装（3 个包）
 ```bash
 npm install @sssxyd/face-liveness-detector @vladmandic/human @techstark/opencv-js
 ```
-Or with yarn:
+<details>
+<summary><strong>其他包管理器</strong></summary>
 ```bash
+# Yarn
 yarn add @sssxyd/face-liveness-detector @vladmandic/human @techstark/opencv-js
-```
-Or with pnpm:
-```bash
+# pnpm
 pnpm add @sssxyd/face-liveness-detector @vladmandic/human @techstark/opencv-js
 ```
-> **Note**: `@vladmandic/human` and `@techstark/opencv-js` are peer dependencies and must be installed separately to avoid bundling large libraries. This keeps your final bundle size smaller if you're already using these libraries elsewhere in your project.
+</details>
+> 📝 **为什么需要三个包？**
+> `@vladmandic/human` 和 `@techstark/opencv-js` 是对等依赖（peer dependencies），需要单独安装以避免捆绑大型库，减小最终的打包体积。
+---
+## ⚠️ 必要配置步骤
-## Quick Start - Using Local Resources
+### 1️⃣ 修复 OpenCV.js ESM 兼容性问题
-> ⚠️ **CRITICAL**: `@techstark/opencv-js` contains an ESM incompatible UMD-format OpenCV.js library that **will cause load failures**. You MUST apply the patch script.
-> - **Issue**: https://github.com/TechStark/opencv-js/issues/44
-> - **Patch Script**: [patch-opencv.js](https://github.com/sssxyd/face-liveness-detector/tree/main/demos/vue-demo/scripts/patch-opencv.js)
-> - **Setup**: Add to your `package.json` scripts as a `postinstall` hook to auto-apply after dependencies install
+`@techstark/opencv-js` 包含不兼容的 UMD 格式，**必须应用补丁脚本**。
-> ⚠️ **CRITICAL**: `@vladmandic/human` requires downloading large model files and TensorFlow WASM backend files. The component **will fail to load without these resources**. Download them to your project directory and configure the paths.
-> - **Models Download Script**: [copy-models.js](https://github.com/sssxyd/face-liveness-detector/tree/main/demos/vue-demo/scripts/copy-models.js)
-> - **WASM Download Script**: [download-wasm.js](https://github.com/sssxyd/face-liveness-detector/tree/main/demos/vue-demo/scripts/download-wasm.js)
-> - **Setup**: Run both scripts as `postinstall` hooks, then configure paths in your engine config
+**参考：**
+- 问题详情：[TechStark/opencv-js#44](https://github.com/TechStark/opencv-js/issues/44)
+- 补丁脚本：[patch-opencv.js](https://github.com/sssxyd/face-liveness-detector/tree/main/demos/vue-demo/scripts/patch-opencv.js)
+**设置方法（推荐）：** 添加到 `package.json` 的 `postinstall` 钩子
+```json
+{
+  "scripts": {
+    "postinstall": "node patch-opencv.cjs"
+  }
+}
+```
+### 2️⃣ 下载 Human.js 模型文件
+`@vladmandic/human` 需要模型文件和 TensorFlow WASM 后端，**否则无法加载**。
+**下载脚本：**
+- 模型复制：[copy-models.js](https://github.com/sssxyd/face-liveness-detector/tree/main/demos/vue-demo/scripts/copy-models.js)
+- WASM 下载：[download-wasm.js](https://github.com/sssxyd/face-liveness-detector/tree/main/demos/vue-demo/scripts/download-wasm.js)
+**设置方法（推荐）：** 配置为 `postinstall` 钩子
+```json
+{
+  "scripts": {
+    "postinstall": "node scripts/copy-models.js && node scripts/download-wasm.js"
+  }
+}
+```
+---
+## 🎯 快速开始
+### 基础示例
 ```typescript
 import FaceDetectionEngine, { LivenessAction } from '@sssxyd/face-liveness-detector'
-// Initialize the engine with custom configuration
+// 初始化引擎
 const engine = new FaceDetectionEngine({
-  // Configure resource paths
+  // 资源路径配置
   human_model_path: '/models',
   tensorflow_wasm_path: '/wasm',
+  tensorflow_backend: 'auto',
-  // Detection settings
-  video_width: 640,
-  video_height: 640,
-  // Quality settings
-  min_image_quality: 0.5,
-  min_face_frontal: 0.9,
-  // Liveness settings - choose your preferred actions
-  liveness_action_count: 1,  // 0 for silent detection only, 1-3 for action-based
-  liveness_action_list: [LivenessAction.BLINK, LivenessAction.MOUTH_OPEN, LivenessAction.NOD]
+  // 检测设置（建议 ≥720p，否则屏幕检测准确率下降）
+  detect_video_ideal_width: 1920,
+  detect_video_ideal_height: 1080,
+  detect_video_mirror: true,
+  detect_video_load_timeout: 5000,
+  detect_frame_delay: 100,
+  // 采集质量要求
+  collect_min_collect_count: 3,        // 最少采集 3 张人脸
+  collect_min_face_ratio: 0.5,         // 人脸占比 50%+
+  collect_max_face_ratio: 0.9,         // 人脸占比 90% 以下
+  collect_min_face_frontal: 0.9,       // 人脸正对度 90%
+  collect_min_image_quality: 0.5,      // 图像质量 50%+
+  // 活体检测设置
+  action_liveness_action_count: 1,     // 需要 1 个动作
+  action_liveness_action_list: [LivenessAction.BLINK, LivenessAction.MOUTH_OPEN, LivenessAction.NOD],
+  action_liveness_action_randomize: true,
+  action_liveness_verify_timeout: 60000,
+  // 防欺骗设置
+  motion_liveness_min_motion_score: 0.15,
+  motion_liveness_strict_photo_detection: false,
+  screen_capture_confidence_threshold: 0.7,
 })
-// Listen for events
+// 监听核心事件
 engine.on('detector-loaded', (data) => {
-  console.log('✅ Engine is ready')
-  console.log(`OpenCV: ${data.opencv_version}`)
-  console.log(`Human.js: ${data.human_version}`)
+  if (data.success) {
+    console.log('✅ 引擎就绪', {
+      opencv: data.opencv_version,
+      human: data.human_version
+    })
+  }
 })
 engine.on('detector-info', (data) => {
-  // Real-time detection information
+  // 每帧实时数据
   console.log({
-    quality: (data.quality * 100).toFixed(1) + '%',
-    frontal: (data.frontal * 100).toFixed(1) + '%',
-    liveness: (data.live * 100).toFixed(1) + '%',
-    realness: (data.real * 100).toFixed(1) + '%'
+    status: data.code,
+    quality: (data.imageQuality * 100).toFixed(1) + '%',
+    frontal: (data.faceFrontal * 100).toFixed(1) + '%',
+    motion: (data.motionScore * 100).toFixed(1) + '%',
+    screen: (data.screenConfidence * 100).toFixed(1) + '%'
   })
 })
 engine.on('detector-action', (data) => {
-  // Action liveness prompts
-  if (data.status === 'started') {
-    console.log(`Please perform: ${data.action}`)
-  } else if (data.status === 'completed') {
-    console.log(`✅ Action recognized: ${data.action}`)
-  }
+  // 动作提示
+  console.log(`请执行动作: ${data.action} (${data.status})`)
 })
 engine.on('detector-finish', (data) => {
+  // 检测完成
   if (data.success) {
-    console.log('✅ Liveness verification passed!')
-    console.log({
-      silentDetections: data.silentPassedCount,
-      actionsCompleted: data.actionPassedCount,
-      imageQuality: (data.bestQualityScore * 100).toFixed(1) + '%',
-      totalTime: (data.totalTime / 1000).toFixed(2) + 's',
-      bestFrame: data.bestFrameImage,  // Base64 encoded
-      bestFace: data.bestFaceImage     // Base64 encoded
+    console.log('✅ 活体验证通过！', {
+      静默通过: data.silentPassedCount,
+      动作完成: data.actionPassedCount,
+      最佳质量: (data.bestQualityScore * 100).toFixed(1) + '%',
+      总耗时: (data.totalTime / 1000).toFixed(2) + 's'
     })
   } else {
-    console.log('❌ Liveness verification failed')
+    console.log('❌ 活体验证失败')
   }
 })
 engine.on('detector-error', (error) => {
-  console.error(`Error [${error.code}]: ${error.message}`)
+  console.error(`❌ 错误 [${error.code}]: ${error.message}`)
 })
-engine.on('detector-debug', (debug) => {
-  console.log(`[${debug.stage}] ${debug.message}`)
-})
-// Initialize and start detection
-async function runDetection() {
+// 启动检测
+async function startLivenessDetection() {
   try {
-    // Initialize libraries (models, TensorFlow WASM, etc.)
+    // 初始化库
     await engine.initialize()
-    // Get video element
-    const videoElement = document.getElementById('video') as HTMLVideoElement
+    // 获取视频元素并开始检测
+    const videoEl = document.getElementById('video') as HTMLVideoElement
+    await engine.startDetection(videoEl)
-    // Start detection on the video stream
-    await engine.startDetection(videoElement)
-    // Detection runs until completion or error
-    // Stop manually if needed:
-    // engine.stopDetection(true)  // true to display best image
+    // 检测自动运行到完成或手动停止
+    // engine.stopDetection(true)  // 停止并显示最佳图像
   } catch (error) {
-    console.error('Failed to run detection:', error)
+    console.error('检测启动失败:', error)
   }
 }
-// Call when ready
-runDetection()
-```
+// 就绪时启动
+startLivenessDetection()
+```
+---
+## ⚙️ 详细配置参考
+### 资源路径配置
+| 选项 | 类型 | 说明 | 默认值 |
+|-----|------|------|--------|
+| `human_model_path` | `string` | Human.js 模型文件目录 | `undefined` |
+| `tensorflow_wasm_path` | `string` | TensorFlow WASM 文件目录 | `undefined` |
+| `tensorflow_backend` | `'auto' \| 'webgl' \| 'wasm'` | TensorFlow 后端引擎 | `'auto'` |
-## Configuration
+### 视频检测设置
-### FaceDetectionEngineConfig
+| 选项 | 类型 | 说明 | 默认值 |
+|-----|------|------|--------|
+| `detect_video_ideal_width` | `number` | 视频宽度（像素） | `1920` |
+| `detect_video_ideal_height` | `number` | 视频高度（像素） | `1080` |
+| `detect_video_mirror` | `boolean` | 水平翻转视频 | `true` |
+| `detect_video_load_timeout` | `number` | 加载超时（ms） | `5000` |
+| `detect_frame_delay` | `number` | 帧间延迟（ms） | `100` |
+| `detect_error_retry_delay` | `number` | 错误重试延迟（ms） | `200` |
-#### Resource Paths
+### 人脸采集质量要求
-| Property | Type | Description | Default |
-|----------|------|-------------|---------|
-| `human_model_path` | `string` | Path to Human.js model files | `undefined` |
-| `tensorflow_wasm_path` | `string` | Path to TensorFlow WASM files | `undefined` |
-| `tensorflow_backend` | `'auto' \| 'webgl' \| 'wasm'` | TensorFlow backend selection | `'auto'` |
+| 选项 | 类型 | 说明 | 默认值 |
+|-----|------|------|--------|
+| `collect_min_collect_count` | `number` | 最少采集数量 | `3` |
+| `collect_min_face_ratio` | `number` | 最小人脸占比 (0-1) | `0.5` |
+| `collect_max_face_ratio` | `number` | 最大人脸占比 (0-1) | `0.9` |
+| `collect_min_face_frontal` | `number` | 最小正对度 (0-1) | `0.9` |
+| `collect_min_image_quality` | `number` | 最小图像质量 (0-1) | `0.5` |
-#### Video Detection Settings
+### 人脸正对度参数
-| Property | Type | Description | Default |
-|----------|------|-------------|---------|
-| `video_width` | `number` | Video stream width in pixels | `640` |
-| `video_height` | `number` | Video stream height in pixels | `640` |
-| `video_mirror` | `boolean` | Mirror video horizontally | `true` |
-| `video_load_timeout` | `number` | Video stream loading timeout in ms | `5000` |
-| `detection_frame_delay` | `number` | Delay between detection frames in ms | `100` |
-| `error_retry_delay` | `number` | Error retry delay in ms | `200` |
+| 选项 | 类型 | 说明 | 默认值 |
+|-----|------|------|--------|
+| `yaw_threshold` | `number` | 偏航角阈值（度） | `3` |
+| `pitch_threshold` | `number` | 俯仰角阈值（度） | `4` |
+| `roll_threshold` | `number` | 翻滚角阈值（度） | `2` |
-#### Detection Quality Settings
+### 图像质量参数
-| Property | Type | Description | Default |
-|----------|------|-------------|---------|
-| `silent_detect_count` | `number` | Number of silent detections to collect | `3` |
-| `min_face_ratio` | `number` | Minimum face size ratio (0-1) | `0.5` |
-| `max_face_ratio` | `number` | Maximum face size ratio (0-1) | `0.9` |
-| `min_face_frontal` | `number` | Minimum face frontality score (0-1) | `0.9` |
-| `min_image_quality` | `number` | Minimum image quality score (0-1) | `0.5` |
-| `min_live_score` | `number` | Minimum liveness score (0-1) | `0.5` |
-| `min_real_score` | `number` | Minimum anti-spoofing score (0-1) | `0.85` |
-| `suspected_frauds_count` | `number` | Number of frauds to detect before fail | `3` |
+| 选项 | 类型 | 说明 | 默认值 |
+|-----|------|------|--------|
+| `require_full_face_in_bounds` | `boolean` | 人脸完全在边界内 | `false` |
+| `min_laplacian_variance` | `number` | 最小模糊检测值 | `40` |
+| `min_gradient_sharpness` | `number` | 最小清晰度 | `0.15` |
+| `min_blur_score` | `number` | 最小模糊分数 | `0.6` |
-#### Face Frontality Features (`face_frontal_features`)
+### 活体检测设置
-| Property | Type | Description | Default |
-|----------|------|-------------|---------|
-| `yaw_threshold` | `number` | Yaw angle threshold in degrees | `3` |
-| `pitch_threshold` | `number` | Pitch angle threshold in degrees | `4` |
-| `roll_threshold` | `number` | Roll angle threshold in degrees | `2` |
+| 选项 | 类型 | 说明 | 默认值 |
+|-----|------|------|--------|
+| `action_liveness_action_list` | `LivenessAction[]` | 动作列表 | `[BLINK, MOUTH_OPEN, NOD]` |
+| `action_liveness_action_count` | `number` | 需要完成的动作数 | `1` |
+| `action_liveness_action_randomize` | `boolean` | 随机化动作顺序 | `true` |
+| `action_liveness_verify_timeout` | `number` | 超时时间（ms） | `60000` |
+| `action_liveness_min_mouth_open_percent` | `number` | 最小张嘴比例 (0-1) | `0.2` |
-#### Image Quality Features (`image_quality_features`)
+### 动作活体检测（防照片攻击）
-| Property | Type | Description | Default |
-|----------|------|-------------|---------|
-| `require_full_face_in_bounds` | `boolean` | Require face completely within bounds | `false` |
-| `use_opencv_enhancement` | `boolean` | Use OpenCV enhancement for quality detection | `true` |
-| `min_laplacian_variance` | `number` | Minimum Laplacian variance for blur detection | `50` |
-| `min_gradient_sharpness` | `number` | Minimum gradient sharpness for blur detection | `0.15` |
-| `min_blur_score` | `number` | Minimum blur score | `0.6` |
+| 选项 | 类型 | 说明 | 默认值 |
+|-----|------|------|--------|
+| `motion_liveness_min_motion_score` | `number` | 最小运动分数 (0-1) | `0.15` |
+| `motion_liveness_min_keypoint_variance` | `number` | 最小关键点方差 (0-1) | `0.02` |
+| `motion_liveness_frame_buffer_size` | `number` | 帧缓冲区大小 | `5` |
+| `motion_liveness_eye_aspect_ratio_threshold` | `number` | 眨眼阈值 | `0.15` |
+| `motion_liveness_motion_consistency_threshold` | `number` | 一致性阈值 (0-1) | `0.3` |
+| `motion_liveness_min_optical_flow_threshold` | `number` | 最小光流幅度 (0-1) | `0.02` |
+| `motion_liveness_strict_photo_detection` | `boolean` | 严格照片检测模式 | `false` |
-#### Liveness Detection Settings
+### 屏幕采集检测
-| Property | Type | Description | Default |
-|----------|------|-------------|---------|
-| `liveness_action_list` | `LivenessAction[]` | List of liveness actions to detect | `[BLINK, MOUTH_OPEN, NOD]` |
-| `liveness_action_count` | `number` | Number of liveness actions to perform | `1` |
-| `liveness_action_randomize` | `boolean` | Randomize liveness actions order | `true` |
-| `liveness_verify_timeout` | `number` | Timeout for liveness verification in ms | `60000` |
-| `min_mouth_open_percent` | `number` | Minimum mouth open percentage (0-1) | `0.2` |
+| 选项 | 类型 | 说明 | 默认值 |
+|-----|------|------|--------|
+| `screen_capture_confidence_threshold` | `number` | 置信度阈值 (0-1) | `0.7` |
+| `screen_capture_detection_strategy` | `string` | 检测策略 | `'adaptive'` |
+| `screen_moire_pattern_threshold` | `number` | 莫尔纹阈值 (0-1) | `0.65` |
+| `screen_moire_pattern_enable_dct` | `boolean` | 启用 DCT 分析 | `true` |
+| `screen_moire_pattern_enable_edge_detection` | `boolean` | 启用边缘检测 | `true` |
+### 屏幕色彩特征
-## API Reference
+| 选项 | 类型 | 说明 | 默认值 |
+|-----|------|------|--------|
+| `screen_color_saturation_threshold` | `number` | 饱和度阈值 (%) | `40` |
+| `screen_color_rgb_correlation_threshold` | `number` | RGB相关性阈值 (0-1) | `0.75` |
+| `screen_color_pixel_entropy_threshold` | `number` | 熵值阈值 (0-8) | `6.5` |
+| `screen_color_gradient_smoothness_threshold` | `number` | 平滑性阈值 (0-1) | `0.7` |
+| `screen_color_confidence_threshold` | `number` | 置信度阈值 (0-1) | `0.65` |
-### Methods
+### 屏幕 RGB 发光检测
+| 选项 | 类型 | 说明 | 默认值 |
+|-----|------|------|--------|
+| `screen_rgb_low_freq_start_percent` | `number` | 低频段开始 (0-1) | `0.15` |
+| `screen_rgb_low_freq_end_percent` | `number` | 低频段结束 (0-1) | `0.35` |
+| `screen_rgb_energy_score_weight` | `number` | 能量权重 | `0.40` |
+| `screen_rgb_asymmetry_score_weight` | `number` | 不同步权重 | `0.40` |
+| `screen_rgb_difference_factor_weight` | `number` | 差异权重 | `0.20` |
+| `screen_rgb_confidence_threshold` | `number` | 置信度阈值 (0-1) | `0.65` |
+---
+## 🛠️ API 方法参考
+### 核心方法
 #### `initialize(): Promise<void>`
-Load and initialize detection libraries. Must be called before using detection.
+加载并初始化检测库。**必须在使用其他功能前调用。**
 ```typescript
 await engine.initialize()
 ```
 #### `startDetection(videoElement): Promise<void>`
-Start face detection on a video element.
+在视频元素上开始人脸检测。
 ```typescript
-const videoElement = document.getElementById('video') as HTMLVideoElement
-await engine.startDetection(videoElement)
+const videoEl = document.getElementById('video') as HTMLVideoElement
+await engine.startDetection(videoEl)
 ```
 #### `stopDetection(success?: boolean): void`
-Stop the detection process.
+停止检测过程。
 ```typescript
-engine.stopDetection(true)  // true to display best image
+engine.stopDetection(true)  // true: 显示最佳检测图像
 ```
 #### `updateConfig(config): void`
-Update configuration during runtime.
+运行时动态更新配置。
 ```typescript
 engine.updateConfig({
-  min_face_ratio: 0.6,
-  liveness_action_count: 2
+  collect_min_face_ratio: 0.6,
+  action_liveness_action_count: 0
 })
 ```
-#### `getConfig(): FaceDetectionEngineConfig`
-Get current configuration.
+#### `getOptions(): FaceDetectionEngineOptions`
+获取当前配置对象。
 ```typescript
-const config = engine.getConfig()
+const config = engine.getOptions()
 ```
-#### `getStatus(): Object`
-Get engine status.
+#### `getEngineState(): EngineState`
+获取引擎当前状态。
 ```typescript
-const { isReady, isDetecting, isInitializing } = engine.getStatus()
-```
-### Events
+const state = engine.getEngineState()
+```
+---
+## 📡 事件系统
+引擎采用 **TypeScript 事件发射器模式**，所有事件都是类型安全的。
+### 事件列表
+<table>
+<tr>
+<td><strong>detector-loaded</strong></td>
+<td>引擎初始化完成</td>
+</tr>
+<tr>
+<td><strong>detector-info</strong></td>
+<td>每帧实时检测数据</td>
+</tr>
+<tr>
+<td><strong>detector-action</strong></td>
+<td>动作活体提示与状态</td>
+</tr>
+<tr>
+<td><strong>detector-finish</strong></td>
+<td>检测完成（成功/失败）</td>
+</tr>
+<tr>
+<td><strong>detector-error</strong></td>
+<td>错误发生时触发</td>
+</tr>
+<tr>
+<td><strong>detector-debug</strong></td>
+<td>调试信息（开发用）</td>
+</tr>
+</table>
+---
+### 📋 detector-loaded
+**引擎初始化完成时触发**
-The engine uses a TypeScript event emitter pattern. All events are type-safe:
-#### `detector-loaded`
-Fired when the engine finishes initialization.
-**Data:**
 ```typescript
 interface DetectorLoadedEventData {
-  success: boolean        // Whether initialization succeeded
-  error?: string          // Error message if any
-  opencv_version?: string // OpenCV.js version
-  human_version?: string  // Human.js version
+  success: boolean        // 初始化是否成功
+  error?: string          // 错误信息（失败时）
+  opencv_version?: string // OpenCV.js 版本号
+  human_version?: string  // Human.js 版本号
 }
 ```
-**Example:**
+**示例：**
 ```typescript
 engine.on('detector-loaded', (data) => {
   if (data.success) {
-    console.log('✅ Engine ready')
-    console.log(`OpenCV: ${data.opencv_version}`)
-    console.log(`Human.js: ${data.human_version}`)
+    console.log('✅ 引擎就绪')
+    console.log(`OpenCV ${data.opencv_version} | Human.js ${data.human_version}`)
   } else {
-    console.error('Engine failed:', data.error)
+    console.error('❌ 初始化失败:', data.error)
   }
 })
 ```
-#### `detector-info`
-Real-time detection information for each frame.
+---
+### 📊 detector-info
+**每帧返回实时检测数据（高频事件）**
-**Data:**
 ```typescript
 interface DetectorInfoEventData {
-  passed: boolean     // Whether silent liveness check passed
-  code: DetectionCode // Detection status code
-  size: number        // Face size ratio (0-1)
-  frontal: number     // Face frontality score (0-1)
-  quality: number     // Image quality score (0-1)
-  real: number        // Anti-spoofing score (0-1)
-  live: number        // Liveness score (0-1)
+  passed: boolean         // 是否通过静默检测
+  code: DetectionCode     // 检测状态码
+  message: string         // 状态消息
+  faceCount: number       // 检测到的人脸数
+  faceRatio: number       // 人脸占比 (0-1)
+  faceFrontal: number     // 人脸正对度 (0-1)
+  imageQuality: number    // 图像质量分数 (0-1)
+  motionScore: number     // 运动分数 (0-1)
+  keypointVariance: number // 关键点方差 (0-1)
+  motionType: string      // 检测到的运动类型
+  screenConfidence: number // 屏幕采集置信度 (0-1)
 }
 ```
-**Detection Codes:**
+**检测状态码：**
 ```typescript
 enum DetectionCode {
-  VIDEO_NO_FACE = 'VIDEO_NO_FACE',        // No face detected
-  MULTIPLE_FACE = 'MULTIPLE_FACE',        // Multiple faces detected
-  FACE_TOO_SMALL = 'FACE_TOO_SMALL',      // Face too small
-  FACE_TOO_LARGE = 'FACE_TOO_LARGE',      // Face too large
-  FACE_NOT_FRONTAL = 'FACE_NOT_FRONTAL',  // Face not frontal enough
-  FACE_NOT_REAL = 'FACE_NOT_REAL',        // Suspected spoofing
-  FACE_NOT_LIVE = 'FACE_NOT_LIVE',        // Low liveness score
-  FACE_LOW_QUALITY = 'FACE_LOW_QUALITY',  // Image quality too low
-  FACE_CHECK_PASS = 'FACE_CHECK_PASS'     // All checks passed
+  VIDEO_NO_FACE = 'VIDEO_NO_FACE',           // 未检测到人脸
+  MULTIPLE_FACE = 'MULTIPLE_FACE',           // 检测到多张人脸
+  FACE_TOO_SMALL = 'FACE_TOO_SMALL',         // 人脸太小
+  FACE_TOO_LARGE = 'FACE_TOO_LARGE',         // 人脸太大
+  FACE_NOT_FRONTAL = 'FACE_NOT_FRONTAL',     // 人脸不够正面
+  FACE_NOT_REAL = 'FACE_NOT_REAL',           // 疑似欺骗
+  FACE_NOT_LIVE = 'FACE_NOT_LIVE',           // 活体分数过低
+  FACE_LOW_QUALITY = 'FACE_LOW_QUALITY',     // 图像质量过低
+  FACE_CHECK_PASS = 'FACE_CHECK_PASS'        // 所有检查通过 ✅
 }
 ```
-**Example:**
+**示例：**
 ```typescript
 engine.on('detector-info', (data) => {
   console.log({
-    passed: data.passed,
-    status: data.code,
-    quality: (data.quality * 100).toFixed(1) + '%',
-    frontal: (data.frontal * 100).toFixed(1) + '%',
-    liveness: (data.live * 100).toFixed(1) + '%',
-    realness: (data.real * 100).toFixed(1) + '%'
+    检测状态: data.code,
+    静默通过: data.passed ? '✅' : '❌',
+    图像质量: `${(data.imageQuality * 100).toFixed(1)}%`,
+    人脸正对度: `${(data.faceFrontal * 100).toFixed(1)}%`,
+    运动分数: `${(data.motionScore * 100).toFixed(1)}%`,
+    屏幕采集: `${(data.screenConfidence * 100).toFixed(1)}%`
   })
 })
 ```
-#### `detector-action`
-Action liveness prompts and status updates.
+---
+### 👤 detector-action
+**动作活体提示与识别状态**
-**Data:**
 ```typescript
 interface DetectorActionEventData {
-  action: LivenessAction    // The action to perform
-  status: LivenessActionStatus // Action status
+  action: LivenessAction          // 要执行的动作
+  status: LivenessActionStatus    // 动作状态
 }
-```
-**Action Types:**
-```typescript
 enum LivenessAction {
-  BLINK = 'blink',
-  MOUTH_OPEN = 'mouth_open',
-  NOD = 'nod'
+  BLINK = 'blink',           // 眨眼
+  MOUTH_OPEN = 'mouth_open', // 张嘴
+  NOD = 'nod'                // 点头
 }
-```
-**Action Status:**
-```typescript
 enum LivenessActionStatus {
-  STARTED = 'started',      // Action prompt started
-  COMPLETED = 'completed',  // Action recognized
-  TIMEOUT = 'timeout'       // Action recognition timeout
+  STARTED = 'started',      // 提示已开始
+  COMPLETED = 'completed',  // 成功识别
+  TIMEOUT = 'timeout'       // 识别超时
 }
 ```
-**Example:**
+**示例：**
 ```typescript
 engine.on('detector-action', (data) => {
+  const actionLabels = {
+    'blink': '眨眼',
+    'mouth_open': '张嘴',
+    'nod': '点头'
+  }
   switch (data.status) {
     case 'started':
-      console.log(`👤 Please perform: ${data.action}`)
-      // Update UI to show action prompt
+      console.log(`👤 请执行: ${actionLabels[data.action]}`)
+      // 显示 UI 提示
       break
     case 'completed':
-      console.log(`✅ Action recognized: ${data.action}`)
-      // Update progress indicator
+      console.log(`✅ 已识别: ${actionLabels[data.action]}`)
+      // 更新进度条
       break
     case 'timeout':
-      console.log(`⏱️ Action timeout: ${data.action}`)
+      console.log(`⏱️ 超时: ${actionLabels[data.action]}`)
+      // 显示重试提示
       break
   }
 })
 ```
-#### `detector-finish`
-Fired when liveness detection completes (successfully or failed).
+---
+### ✅ detector-finish
+**检测流程完成（成功或失败）**
-**Data:**
 ```typescript
 interface DetectorFinishEventData {
-  success: boolean         // Whether liveness verification passed
-  silentPassedCount: number    // Number of silent detections passed
-  actionPassedCount: number    // Number of actions completed
-  totalTime: number        // Total detection time in ms
-  bestQualityScore: number // Best image quality score (0-1)
-  bestFrameImage: string | null  // Base64 encoded best frame image
-  bestFaceImage: string | null   // Base64 encoded best face image
+  success: boolean         // 是否通过验证
+  silentPassedCount: number    // 静默检测通过数
+  actionPassedCount: number    // 动作完成数
+  totalTime: number        // 总耗时（毫秒）
+  bestQualityScore: number // 最佳图像质量 (0-1)
+  bestFrameImage: string | null  // Base64 帧图像
+  bestFaceImage: string | null   // Base64 人脸图像
 }
 ```
-**Example:**
+**示例：**
 ```typescript
 engine.on('detector-finish', (data) => {
   if (data.success) {
-    console.log('✅ Liveness verification passed!')
-    console.log({
-      silentDetections: data.silentPassedCount,
-      actionsCompleted: data.actionPassedCount,
-      quality: (data.bestQualityScore * 100).toFixed(1) + '%',
-      time: (data.totalTime / 1000).toFixed(2) + 's'
+    console.log('🎉 活体验证成功！', {
+      静默通过: `${data.silentPassedCount} 次`,
+      动作完成: `${data.actionPassedCount} 次`,
+      最佳质量: `${(data.bestQualityScore * 100).toFixed(1)}%`,
+      总耗时: `${(data.totalTime / 1000).toFixed(2)}s`
     })
-    // Send results to server
+    // 上传结果到服务器
     if (data.bestFrameImage) {
-      uploadVerificationResult({
+      uploadToServer({
         image: data.bestFrameImage,
         quality: data.bestQualityScore,
         timestamp: new Date()
       })
     }
   } else {
-    console.log('❌ Liveness verification failed')
-    // Prompt user to try again
+    console.log('❌ 验证失败，请重试')
   }
 })
 ```
-#### `detector-error`
-Fired when an error occurs during detection.
+---
+### ⚠️ detector-error
+**检测过程中发生错误**
-**Data:**
 ```typescript
 interface DetectorErrorEventData {
-  code: ErrorCode // Error code
-  message: string // Error message
+  code: ErrorCode  // 错误代码
+  message: string  // 错误信息
 }
-```
-**Error Codes:**
-```typescript
 enum ErrorCode {
   DETECTOR_NOT_INITIALIZED = 'DETECTOR_NOT_INITIALIZED',
   CAMERA_ACCESS_DENIED = 'CAMERA_ACCESS_DENIED',
@@ -448,124 +612,118 @@ enum ErrorCode {
 }
 ```
-**Example:**
+**示例：**
 ```typescript
 engine.on('detector-error', (error) => {
-  console.error(`❌ Error [${error.code}]: ${error.message}`)
-  switch (error.code) {
-    case 'CAMERA_ACCESS_DENIED':
-      showErrorMessage('Please grant camera permissions')
-      break
-    case 'STREAM_ACQUISITION_FAILED':
-      showErrorMessage('Failed to access camera')
-      break
-    case 'SUSPECTED_FRAUDS_DETECTED':
-      showErrorMessage('Spoofing detected - please try again')
-      break
-    default:
-      showErrorMessage('Detection failed: ' + error.message)
+  const errorMessages: Record<string, string> = {
+    'DETECTOR_NOT_INITIALIZED': '引擎未初始化',
+    'CAMERA_ACCESS_DENIED': '摄像头权限被拒绝',
+    'STREAM_ACQUISITION_FAILED': '无法获取摄像头数据流',
+    'SUSPECTED_FRAUDS_DETECTED': '检测到欺骗行为'
   }
+  console.error(`❌ 错误 [${error.code}]: ${errorMessages[error.code] || error.message}`)
+  showUserErrorPrompt(errorMessages[error.code])
 })
 ```
-#### `detector-debug`
-Debug information for development and troubleshooting.
+---
+### 🐛 detector-debug
+**开发和故障排除的调试信息**
-**Data:**
 ```typescript
 interface DetectorDebugEventData {
-  level: 'info' | 'warn' | 'error'  // Debug level
-  stage: string                      // Current processing stage
-  message: string                    // Debug message
-  details?: Record<string, any>      // Additional details
-  timestamp: number                  // Unix timestamp
+  level: 'info' | 'warn' | 'error'  // 日志级别
+  stage: string                      // 处理阶段
+  message: string                    // 调试信息
+  details?: Record<string, any>      // 额外详情
+  timestamp: number                  // Unix 时间戳
 }
 ```
-**Example:**
+**示例：**
 ```typescript
 engine.on('detector-debug', (debug) => {
   const time = new Date(debug.timestamp).toLocaleTimeString()
-  console.log(`[${time}] [${debug.stage}] ${debug.message}`)
+  const prefix = `[${time}] [${debug.stage}]`
-  if (debug.details) {
-    console.log('Details:', debug.details)
-  }
-  // Log errors for troubleshooting
   if (debug.level === 'error') {
-    logErrorToServer({
-      stage: debug.stage,
-      message: debug.message,
-      details: debug.details
-    })
+    console.error(`${prefix} ❌ ${debug.message}`, debug.details)
+  } else {
+    console.log(`${prefix} ℹ️ ${debug.message}`)
   }
 })
 ```
-## Enumerations
+---
+## 📖 类型定义
 ### LivenessAction
 ```typescript
 enum LivenessAction {
-  BLINK = 'blink',
-  MOUTH_OPEN = 'mouth_open',
-  NOD = 'nod'
+  BLINK = 'blink',           // 眨眼
+  MOUTH_OPEN = 'mouth_open', // 张嘴
+  NOD = 'nod'                // 点头
 }
 ```
 ### LivenessActionStatus
 ```typescript
 enum LivenessActionStatus {
-  STARTED = 'started',      // Action prompt has started
-  COMPLETED = 'completed',  // Action successfully recognized
-  TIMEOUT = 'timeout'       // Action recognition timeout
+  STARTED = 'started',      // 动作提示已开始
+  COMPLETED = 'completed',  // 动作成功识别
+  TIMEOUT = 'timeout'       // 动作识别超时
 }
 ```
 ### DetectionCode
 ```typescript
 enum DetectionCode {
-  VIDEO_NO_FACE = 'VIDEO_NO_FACE',            // No face detected in the video
-  MULTIPLE_FACE = 'MULTIPLE_FACE',            // Multiple faces detected
-  FACE_TOO_SMALL = 'FACE_TOO_SMALL',          // Face size below minimum threshold
-  FACE_TOO_LARGE = 'FACE_TOO_LARGE',          // Face size above maximum threshold
-  FACE_NOT_FRONTAL = 'FACE_NOT_FRONTAL',      // Face angle not frontal enough
-  FACE_NOT_REAL = 'FACE_NOT_REAL',            // Suspected spoofing detected
-  FACE_NOT_LIVE = 'FACE_NOT_LIVE',            // Liveness score below threshold
-  FACE_LOW_QUALITY = 'FACE_LOW_QUALITY',      // Image quality below minimum
-  FACE_CHECK_PASS = 'FACE_CHECK_PASS'         // All detection checks passed
+  VIDEO_NO_FACE = 'VIDEO_NO_FACE',           // 视频中未检测到人脸
+  MULTIPLE_FACE = 'MULTIPLE_FACE',           // 检测到多张人脸
+  FACE_TOO_SMALL = 'FACE_TOO_SMALL',         // 人脸尺寸小于最小阈值
+  FACE_TOO_LARGE = 'FACE_TOO_LARGE',         // 人脸尺寸大于最大阈值
+  FACE_NOT_FRONTAL = 'FACE_NOT_FRONTAL',     // 人脸角度不够正面
+  FACE_NOT_REAL = 'FACE_NOT_REAL',           // 检测到疑似欺骗
+  FACE_NOT_LIVE = 'FACE_NOT_LIVE',           // 活体分数低于阈值
+  FACE_LOW_QUALITY = 'FACE_LOW_QUALITY',     // 图像质量低于最小值
+  FACE_CHECK_PASS = 'FACE_CHECK_PASS'        // 所有检测检查通过 ✅
 }
 ```
 ### ErrorCode
 ```typescript
 enum ErrorCode {
-  DETECTOR_NOT_INITIALIZED = 'DETECTOR_NOT_INITIALIZED',  // Engine not initialized
-  CAMERA_ACCESS_DENIED = 'CAMERA_ACCESS_DENIED',          // Camera permission denied
-  STREAM_ACQUISITION_FAILED = 'STREAM_ACQUISITION_FAILED', // Failed to get video stream
-  SUSPECTED_FRAUDS_DETECTED = 'SUSPECTED_FRAUDS_DETECTED'  // Spoofing/fraud suspected
+  DETECTOR_NOT_INITIALIZED = 'DETECTOR_NOT_INITIALIZED',  // 引擎未初始化
+  CAMERA_ACCESS_DENIED = 'CAMERA_ACCESS_DENIED',          // 摄像头权限被拒
+  STREAM_ACQUISITION_FAILED = 'STREAM_ACQUISITION_FAILED', // 获取视频流失败
+  SUSPECTED_FRAUDS_DETECTED = 'SUSPECTED_FRAUDS_DETECTED'  // 检测到欺骗/欺诈
 }
 ```
-## Advanced Usage
+---
-For comprehensive examples and advanced usage patterns, please refer to the official demo project:
+## 🎓 高级用法与示例
-**👉 [Vue Demo Project](https://github.com/sssxyd/face-liveness-detector/tree/main/demos/vue-demo/)**
+### 完整的 Vue 3 演示项目
-The demo includes:
-- Complete Vue 3 integration with TypeScript
-- Real-time detection visualization
-- Configuration panel for experimenting with different settings
-- Event handling examples for all engine events
-- Debug panel showing detailed detection information
-- Responsive UI design for mobile and desktop
-- Error handling and user feedback patterns
-- Result export and image capture examples
+有关全面的示例和高级使用模式，请参考官方演示项目：
-To run the demo locally:
+**[Vue 演示项目](https://github.com/sssxyd/face-liveness-detector/tree/main/demos/vue-demo/)** 包括：
+- ✅ 完整的 Vue 3 + TypeScript 集成
+- ✅ 实时检测结果可视化
+- ✅ 动态配置面板
+- ✅ 所有引擎事件的完整处理
+- ✅ 实时调试面板
+- ✅ 响应式移动端 + 桌面端 UI
+- ✅ 错误处理和用户反馈
+- ✅ 结果导出和图像捕获
+**快速启动演示：**
 ```bash
 cd demos/vue-demo
@@ -573,123 +731,116 @@ npm install
 npm run dev
 ```
-Then open your browser to the displayed local URL to see the detection engine in action.
+然后在浏览器中打开显示的本地 URL。
+---
-## Downloading and Hosting Model Files
+## 📥 本地部署模型文件
-To avoid CDN dependencies and improve performance, you can download model files locally:
+### 为什么需要本地部署？
-### Available Download Scripts
+- 🚀 **提升性能** - 避免 CDN 延迟
+- 🔒 **隐私保护** - 完全离线运行
+- 🌐 **网络独立** - 不依赖外部连接
-Two scripts are provided in the root directory:
+### 可用脚本
-#### 1. Copy Human.js Models
+项目根目录提供两个下载脚本：
+#### 1️⃣ 复制 Human.js 模型
 ```bash
-node copy-human-models.js
+node copy-models.js
 ```
-**What it does:**
-- Copies face detection models from `node_modules/@vladmandic/human/models`
-- Saves to `public/models/` directory
-- Downloads both `.json` and `.bin` model files
-- Shows file size and progress
+**功能：**
+- 从 `node_modules/@vladmandic/human/models` 复制模型
+- 保存到 `public/models/` 目录
+- 包含 `.json` 和 `.bin` 模型文件
+- 自动显示文件大小和进度
-#### 2. Download TensorFlow.js WASM Files
+#### 2️⃣ 下载 TensorFlow WASM 文件
 ```bash
-node download-tensorflow-wasm.js
+node download-wasm.js
 ```
-**What it does:**
-- Downloads TensorFlow.js WASM backend files
-- Saves to `public/wasm/` directory
-- Downloads 4 critical files:
+**功能：**
+- 自动下载 TensorFlow.js WASM 后端
+- 保存到 `public/wasm/` 目录
+- 下载 4 个关键文件：
   - `tf-backend-wasm.min.js`
   - `tfjs-backend-wasm.wasm`
   - `tfjs-backend-wasm-simd.wasm`
   - `tfjs-backend-wasm-threaded-simd.wasm`
-- **Supports multiple CDN sources** with automatic fallback:
-  1. unpkg.com (primary)
-  2. cdn.jsdelivr.net (backup)
-  3. esm.sh (fallback)
-  4. cdn.esm.sh (last resort)
+- **智能多 CDN 源** 自动回退：
+  1. unpkg.com（推荐）
+  2. cdn.jsdelivr.net
+  3. esm.sh
+  4. cdn.esm.sh
-### Configuration to Use Local Files
+### 配置项目使用本地文件
-Once downloaded, configure the engine to use these local files:
+下载完成后，在引擎初始化时指定本地路径：
 ```typescript
 const engine = new FaceDetectionEngine({
-  // Use local files instead of CDN
+  // 使用本地文件而不是 CDN
   human_model_path: '/models',
   tensorflow_wasm_path: '/wasm',
-  // ... rest of configuration
+  // 其他配置...
 })
 ```
-## Browser Requirements
+### 自动化设置（推荐）
+在 `package.json` 中配置 `postinstall` 钩子实现自动下载：
+```json
+{
+  "scripts": {
+    "postinstall": "node scripts/copy-models.js && node scripts/download-wasm.js"
+  }
+}
+```
+---
+## 🌐 浏览器兼容性
-- Modern browsers with WebRTC support (Chrome, Firefox, Edge, Safari 11+)
-- HTTPS required for getUserMedia
-- WebGL or WASM backend support
+| 浏览器 | 版本 | 支持 | 备注 |
+|--------|------|------|------|
+| Chrome | 60+ | ✅ | 完全支持 |
+| Firefox | 55+ | ✅ | 完全支持 |
+| Safari | 11+ | ✅ | 完全支持 |
+| Edge | 79+ | ✅ | 完全支持 |
-## Performance Tips
+**系统要求：**
-1. **Adjust detection frame delay** - Higher delay = lower CPU usage but slower detection
-   ```typescript
-   engine.updateConfig({ detection_frame_delay: 200 })
-   ```
+- 📱 支持 **WebRTC** 的现代浏览器
+- 🔒 **HTTPS 环境**（开发可用 localhost）
+- ⚙️ **WebGL** 或 **WASM** 后端支持
+- 📹 **用户授权** - 需要摄像头权限
-2. **Reduce canvas size** - Smaller canvases process faster
-   ```typescript
-   engine.updateConfig({
-     video_width: 480,
-     video_height: 480
-   })
-   ```
+---
-3. **Optimize light conditions** - Better lighting = better detection
-   - Avoid backlighting
-   - Ensure face is well-lit
+## 📄 许可证
-4. **Monitor debug output** - Use debug events to identify bottlenecks
-   ```typescript
-   engine.on('detector-debug', (debug) => {
-     if (debug.stage === 'detection') {
-       console.time(debug.message)
-     }
-   })
-   ```
+[MIT License](./LICENSE) - 自由使用和修改
-## Troubleshooting
+## 🤝 贡献
-### "Camera access denied"
-- Ensure HTTPS is used (or localhost for development)
-- Check browser permissions
-- User must grant camera access
+欢迎提交 Issue 和 Pull Request！
-### "Video loading timeout"
-- Check internet connection
-- Verify model files are accessible
-- Increase `video_load_timeout`
+---
-### Poor detection accuracy
-- Ensure good lighting
-- Keep face centered in frame
-- Face should be 50-90% of frame
-- Face should be frontal (not tilted)
+<div align="center">
-### High CPU usage
-- Increase `detection_frame_delay`
-- Reduce `video_width` and `video_height`
-- Disable `show_action_prompt` if not needed
+**[⬆ 返回顶部](#人脸活体检测引擎)**
-## License
+Made with ❤️ by [sssxyd](https://github.com/sssxyd)
-MIT
+</div>
-## Support
-For issues and questions, please visit: https://github.com/sssxyd/face-liveness-detector/issues