This code is built off Sascha Dittman's object detection with TensorFlow.JS demo: https://github.com/SaschaDittmann/tfjs-cv-objectdetection. The model is a YOLOv8n model, converted to the TensorFlow.JS format with this tutorial: https://docs.ultralytics.com/modes/export/#arguments. The code to convert raw model output into coherent bounding boxes comes from https://github.com/Hyuto/yolov8-tfjs/. The model is trained off the COCO dataset, so its just for general object detection.
This model loads and runs far faster than its contemporaries and is very small.
Prepare the node environments:
$ yarnRun the local web server script:
$ node server.jsThen, to get a URL to the port this has ran the server on, install localtunnel and use it:
$ npm install -g localtunnel
$ lt --port 3000