TypeError: trackerFn(.).registerTensor不是函数

Question

这里是我的设置：

• node -v :v12.8.1
• Python3.7.1
• System版本: macOS 10.14.1 (18B75)

Package.json:

"@tensorflow-models/mobilenet": "^2.0.4",
"@tensorflow/tfjs": "^1.6.0",
"@tensorflow/tfjs-node": "^1.2.11",

注意:必须更改为这些设置才能解决this issue.

我的文件架构：

​

​

**我的代码：**

const readImage = path => {
  const imageBuffer = fs.readFileSync(path);
  const tfimage = tfnode.node.decodeImage(imageBuffer);
  return tfimage;
}


const performTheImageClassification = async path  => {
    const image = readImage("img_to_predict/actual.jpeg");
     const mobilenetModel = await mobilenet.load();
     const predictions = await mobilenetModel.classify(image);
     console.log('Classification Results:', predictions);
}

终端输出：

 (node:65191) UnhandledPromiseRejectionWarning: TypeError: trackerFn(...).registerTensor is not a function
    at new Tensor (/Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/node_modules/@tensorflow/tfjs-node/node_modules/@tensorflow/tfjs-core/dist/tensor.js:246:21)
    at Function.Tensor.make (/Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/node_modules/@tensorflow/tfjs-node/node_modules/@tensorflow/tfjs-core/dist/tensor.js:261:16)
    at makeTensor (/Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/node_modules/@tensorflow/tfjs-node/node_modules/@tensorflow/tfjs-core/dist/ops/tensor_ops.js:98:28)
    at Object.scalar (/Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/node_modules/@tensorflow/tfjs-node/node_modules/@tensorflow/tfjs-core/dist/ops/tensor_ops.js:127:12)
    at NodeJSKernelBackend.decodeJpeg (/Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/node_modules/@tensorflow/tfjs-node/dist/nodejs_kernel_backend.js:1513:38)
    at /Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/node_modules/@tensorflow/tfjs-node/dist/image.js:103:14
    at /Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/node_modules/@tensorflow/tfjs/node_modules/@tensorflow/tfjs-core/dist/engine.js:388:22
    at Engine.scopedRun (/Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/node_modules/@tensorflow/tfjs/node_modules/@tensorflow/tfjs-core/dist/engine.js:398:23)
    at Engine.tidy (/Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/node_modules/@tensorflow/tfjs/node_modules/@tensorflow/tfjs-core/dist/engine.js:387:21)
    at Object.tidy (/Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/node_modules/@tensorflow/tfjs-node/node_modules/@tensorflow/tfjs-core/dist/globals.js:172:28)
    at decodeJpeg (/Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/node_modules/@tensorflow/tfjs-node/dist/image.js:101:24)
    at Object.decodeImage (/Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/node_modules/@tensorflow/tfjs-node/dist/image.js:209:20)
    at readImage (/Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/server.js:97:31)
    at performTheImageClassification (/Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/server.js:88:19)
    at client.search.then.then (/Users/cyrus/Documents/Code/01. Code/debord_slow_sender_meme/server.js:81:12)
    at process._tickCallback (internal/process/next_tick.js:68:7)

Answer 1

我也犯了同样的错误，当我试图修复这个错误的时候，其他的错误也出现了，这个错误还在继续。最后，我有了3个不同版本的tfjs tfjs-节点和移动网络。

我认为这个错误肯定是由版本错配引起的。我将tensorflow文件夹移除在nodemodules文件夹下，然后运行

npm i @tensorflow/tfjs @tensorflow/tfjs-node

我特别指出了这一点，因为首先，我试图通过运行npm卸载命令并重新安装它们来卸载这两个包，但在nodemodules世界中，事情显然比这复杂得多，所以您需要卸载整个tensorflow包，因为我有一个完全不同版本的tfjs-converter，，另一个版本的tfjs--核心，tfjs--数据，以及它真的混在一起，所以确保您删除了整个tensorflow包并运行上面编写的命令来安装必要的包。

这将自动安装两个软件包的最新版本，当我发布这个答案时，它们应该是1.7.4。现在它对我来说是完美无缺的，我希望它也能对你有用。

，这是我的package.json，如果你想确定的话：

"@tensorflow-models/mobilenet": "^2.0.4",
"@tensorflow/tfjs": "^1.7.4",
"@tensorflow/tfjs-node": "^1.7.4",