tensorflow-yolov3

Why we included this project

This codebase is a readable, script-by-script walk through training YOLOv3 rather than just running it. The README covers the whole flow in plain Python: converting the released COCO weights to a checkpoint, freezing a graph for inference, running the image and video demos, then fine-tuning on your own data using a text file of bounding boxes and a class names list. The VOC training path is documented end to end, including how to lay out the dataset directory and which config fields to edit, so you rarely have to guess between steps. Teams comfortable with TensorFlow 1.x who want a reference implementation they can actually read rather than a heavyweight framework will get a working companion to the YOLOv3 paper out of it. One caveat: the project stays on the older TF stack, and the author points to a separate TF2.0 implementation in the repo docs.