supervision is a Python package with tools for building computer vision applications. With supervision, you can:
- Load and filter predictions from a range of state of the art models into a standard API (i.e. YOLOv8, SAM).
- Annotate images and videos with bounding boxes, segmentation masks, and more.
- Calculate confusion matrices.
- Use ByteTrack to track objects.
- Use SAHI to process small object detection.
- Process video frames.
- And more.
Our goal with supervision is to provide model-agnostic tools with a concise syntax that you can use to build logic on top of computer vision models. With supervision, you don't need to write your own bounding box drawing logic, mAP calculator, predictions filtering lgoic, and more. You can use the same API across projects and models.
We are "eating what we cook" while building supervision per se: our team uses supervision every day to solve vision problems and provides feedback to help refine the package.
Try it out and share what you think!
