Supervision
Python toolkit for drawing, tracking, and managing computer vision detections
What it is
Supervision is a Python library that sits around any detection, segmentation, or classification model and handles the surrounding work: converting model outputs into a common detections format, drawing boxes, masks, and labels on frames, tracking objects across video, and counting objects in zones. It also loads, splits, merges, and converts datasets between COCO, YOLO, and Pascal VOC formats. Developers building vision applications use it to avoid rewriting the same glue code for each project, with connectors for Ultralytics, Transformers, MMDetection, and Roboflow Inference.