AS-One

Why we included this project

AS-One wraps several YOLO versions and tracking algorithms behind one class, so you can switch detectors or trackers by changing a flag instead of rewriting your pipeline. That makes it a decent fit for video analytics, traffic monitoring, or retail counting, where you often want to test ByteTrack against DeepSORT or NorFair without much glue code. Models are available in ONNX, PyTorch, and CoreML, so the same code can run on CPU, GPU, or macOS. The examples get a working detector or tracker up in about ten lines, and you can point it at your own trained weights by passing a file path. It also includes drawing helpers and demo scripts, which is enough to prototype with before you move to a heavier stack.