为了防止狗上沙发，写了一个浏览器实时识别目标功能

@派大星

发布于 2024-03-20 21:40:47

960

发布于 2024-03-20 21:40:47

文章被收录于专栏：码上遇见你

给大家分享一个我在网上很有意思的小程序。需求源于生活

作者：xiaowoli 原文链接：https://v2ex.com/t/1024038#reply24

该功能室网页中调用摄像头进行物体识别后可实现的行为

背景

网友的家里有一条狗🐶，很喜欢乘人不备睡沙发🛋️，恰好最近刚搬家 + 狗迎来了掉毛期不想让沙发上很多毛。所以希望能识别到狗，然后播放“gun 下去”的音频📣。

需求分析

需要一个摄像头📷
- 利用 chrome 浏览器可以调用手机摄像头，获取权限，然后利用 video 将摄像头的内容绘制到 video 上。
通过摄像头实时识别画面中的狗🐶
- 利用 tensorflow 和预训练的 COCO-SSD MobileNet V2 模型进行对象检测。
- 将摄像头的视频流转化成视频帧图像传给模型进行识别
录制一个音频
- 识别到目标（狗）后播放音频📣
需要部署在一个设备上
- 找一个不用的旧手机📱，Android 系统
- 安装 termux 来实现开启本地 http 服务🌐

技术要点

利用浏览器 API 调用手机摄像头，将视频流推给 video

const stream = await navigator.mediaDevices.getUserMedia({
  // video: { facingMode: "environment" },  // 摄像头后置
  video: { facingMode: "user" },
});

const videoElement = document.getElementById("camera-stream");
videoElement.srcObject = stream;

加载模型，实现识别

let dogDetector;

async function loadDogDetector() {
  // 加载预训练的 SSD MobileNet V2 模型
  const model = await cocoSsd.load();
  dogDetector = model; // 将加载好的模型赋值给 dogDetector 变量
}

监听 video 的播放，将视频流转换成图像传入模型检测

videoElement.addEventListener("play", async () => {
  requestAnimationFrame(processVideoFrame);
});

async function processVideoFrame() {
  if (!videoElement.paused && !videoElement.ended) {
    canvas.width = videoElement.videoWidth;
    canvas.height = videoElement.videoHeight;
    ctx.drawImage(videoElement, 0, 0, canvas.width, canvas.height);

    // 获取当前帧图像数据
    const imageData = ctx.getImageData(0, 0, canvas.width, canvas.height);

    // 对帧执行预测
    let predictionClasses = "";
    const predictions = await dogDetector.detect(imageData);
    // 处理预测结果，比如检查是否有狗被检测到
    for (const prediction of predictions) {
      predictionClasses += `${prediction.class}\n`; // 组装识别的物体名称
      if (prediction.class === "dog") {
        // 播放声音
        playDogBarkSound();
      }
    }
    nameContainer.innerText = predictionClasses.trim(); // 移除末尾的换行符

    requestAnimationFrame(processVideoFrame);
  }
}

播放音频

async function playDogBarkSound() {
  if (playing) return;
  playing = true;
  const audio = new Audio(dogBarkSound);
  audio.addEventListener("ended", () => {
    playing = false;
  });
  audio.volume = 0.5; // 调整音量大小
  await audio.play();
}

手机开启本地 http 服务

安装 termux
安装 python3
运行 python3 -m http.server 8000

直接用 termux 打开文件
访问 http://localhost:8000

项目代码(改为 html 文件后)


<!DOCTYPE html>
<html lang="en">
  <head>
    <meta charset="UTF-8" />
    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
    <title>Mobile Dog Detector</title>
    <script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs@4.17.0/dist/tf.min.js"></script>
    <script src="https://cdn.jsdelivr.net/npm/@tensorflow-models/coco-ssd@2.2.3/dist/coco-ssd.min.js"></script>
    <style>
      #camera-stream {
        width: 200px;
        height: auto;
      }
      #name {
        height: 200px;
        overflow-y: auto;
        font-family: Arial, sans-serif;
      }
    </style>
  </head>
  <body>
    <video id="camera-stream" autoplay playsinline></video>
    <div id="name" style="height: 200px"></div>

    <script>
      let playing = false;
      let dogDetector;

      async function loadDogDetector() {
        // 加载预训练的 SSD MobileNet V2 模型
        const model = await cocoSsd.load();
        dogDetector = model; // 将加载好的模型赋值给 dogDetector 变量
        console.log("dogDetector", dogDetector);
        startCamera();
      }
      // 调用函数加载模型
      loadDogDetector();

      async function startCamera() {
        const stream = await navigator.mediaDevices.getUserMedia({
          // video: { facingMode: "environment" },  // 摄像头后置
          video: { facingMode: "user" },
        });
        const nameContainer = document.getElementById("name");
        const videoElement = document.getElementById("camera-stream");
        videoElement.srcObject = stream;

        const canvas = document.createElement("canvas");
        const ctx = canvas.getContext("2d");

        videoElement.addEventListener("play", async () => {
          requestAnimationFrame(processVideoFrame);
        });
        async function processVideoFrame() {
          if (!videoElement.paused && !videoElement.ended) {
            canvas.width = videoElement.videoWidth;
            canvas.height = videoElement.videoHeight;
            ctx.drawImage(videoElement, 0, 0, canvas.width, canvas.height);

            const imageData = ctx.getImageData(
              0,
              0,
              canvas.width,
              canvas.height
            );

            let predictionClasses = "";
            const predictions = await dogDetector.detect(imageData);
            for (const prediction of predictions) {
              predictionClasses += `${prediction.class}\n`;
              if (prediction.class === "dog") {
                // 修改为检测到狗时播放声音
                playDogBarkSound();
              }
            }
            nameContainer.innerText = predictionClasses.trim();

            requestAnimationFrame(processVideoFrame);
          }
        }

        async function playDogBarkSound() {
          if (playing) return;
          playing = true;
          const audio = new Audio("./getout.mp3");
          audio.addEventListener("ended", () => {
            playing = false;
          });
          audio.volume = 0.5; // 调整音量大小
          await audio.play();
        }
      }
    </script>
  </body>
</html>