This is adapted and rewritten version of YOLOv8 object segmentation (powered by onnx). This version can be run on JavaScript without any frameworks and demonstrates object detection using web camera.

Setup

To see it at work, just run index.html file.

Models

Main Model

YOLOv8n model converted to onnx with input dimensions of 416x416.

used model : yolov8n.onnx
size       : ~ 12.5Mb

NMS

ONNX model to perform NMS operator [CUSTOM].

nms-yolov8.onnx

Use another model

It is possible to use bigger models converted to onnx, however this might impact the total loading time.

To use another YOLOv8 model, download it from Ultralytics and convert it to onnx file format.

Custom YOLOv8 Object Detection Models

Please update labels object inside of main.js file.

Updates

• Added another web camera based example for YOLOv8 running without any frameworks. In this example there is no need for NMS operator, but it is slower. See: "yolov8_onnx_without_nms" folder.
• Added TFJS version of YOLOv8 which is faster and more robust. See: "tfjs_version" folder.

Demo

To see demo, please visit the following page. To see TFJS version, please visit this page.