This repo uses show how machine learning works directly in the web browser via TensorFlow.js to classify and detect common objects using the Object detection model (coco-ssd) from a live webcam stream. This is a sample vue 3 + vite project.
It actually gives the coordinates of the bounding box for each object which I was able to highlight using #canvas in the actual video stream.
VSCode + Volar (and disable Vetur) + TypeScript Vue Plugin (Volar).
See Vite Configuration Reference.
npm install
npm run dev
npm run build