yolov3-tf2

Why we included this project

This TensorFlow 2 implementation of YOLOv3 is a clean, readable reference rather than a production package. It comes with conversion scripts and pre-trained Darknet weights for both the full model and the faster tiny variant, so you can run detection on images, videos, or a live webcam within minutes. The repo also demonstrates two training paths, eager gradient tape and the standard model.fit workflow, which is helpful if you plan to fine-tune the model on your own classes. A TensorFlow Serving export is included for pushing a trained model into a service. Just expect to adapt the code to newer TF releases, since the project is a learning base more than an actively maintained library.