# roboflow/supervision

We write your reusable computer vision tools. 💜

Repository: https://github.com/roboflow/supervision
Canonical: https://ross.abutalabs.com/products/supervision
Homepage: https://supervision.roboflow.com
Language: Python
License: MIT
License Family: permissive
Topics: computer-vision, image-processing, python, yolo, instance-segmentation, object-detection, tracking, video-processing, coco, pascal-voc, deep-learning, metrics, machine-learning, pytorch, tensorflow, classification, oriented-bounding-box, low-code, hacktoberfest
Last push: 2026-08-25T15:34:03+00:00

## Health v2 (maintenance only)
Score: 95/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 87, longevity 98
- inputs: {"age_days": 1374, "days_push": 8, "days_rel": 9, "gap_med": 41, "n_releases_24m": 12}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 49745, forks 4726 (observed 2026-08-28T04:12:11.663586+00:00)

## What it is
Supervision is a Python library of reusable computer vision tools that bridges the gap between detection/segmentation/classification models and application code. It provides model-agnostic connectors, annotators for drawing boxes/masks/trails, dataset loaders for formats like COCO and Pascal VOC, and utilities for tracking, zone counting, and evaluation metrics.

## Use cases
- draw bounding boxes and masks on detection results
- count objects crossing a line or inside a zone in video
- load and convert COCO or Pascal VOC datasets
- evaluate object detection models with mAP metrics
- track objects across video frames
- plug any detector model into a shared detection API
- annotate and visualize model predictions in notebooks

## When to choose
- you need post-processing, annotation, or evaluation utilities around existing detection or segmentation models
- you want model-agnostic connectors so you can swap YOLO, Transformers, or Roboflow models without rewriting code
- you are building video analytics like zone counting, line crossing, or object tracking on top of detections
- you work in Python notebooks and want quick visualization of CV model outputs

## When to avoid
- you need to train models from scratch - supervision focuses on inference-time tooling, not training
- you need a full end-to-end CV platform with labeling and deployment - consider the broader Roboflow ecosystem instead
- your project is not in Python or requires GPU-level optimized inference itself rather than utilities around model outputs

## Facets
- artifact type: library
- maturity: active
- function: computer-vision, image-processing, video-processing, data-visualization, machine-learning, sdk
- domain: computer-vision, image-processing, machine-learning, deep-learning, data-science
- platform: python, cross-platform
- tags: object-detection, instance-segmentation, object-tracking, yolo, coco, pascal-voc, annotators, zone-counting, model-agnostic, pytorch, tensorflow, detection-metrics, dataset-loading, bounding-boxes, low-code, video

## Member repositories
- roboflow/supervision (main) score 95

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:12:11.663586+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T16:20:11.195602+00:00, confidence not recorded.
  - readme: https://github.com/roboflow/supervision (fetched 2026-08-28T04:12:11.663586+00:00, sha 875d6caa75d5)
  - homepage: https://supervision.roboflow.com (fetched 2026-08-28T18:12:59.195204+00:00, sha 36c6c3c2e4f9)
  - registry_pypi: https://pypi.org/pypi/supervision/json (fetched 2026-08-28T18:12:59.203539+00:00, sha 4be8a6c98c3c)
- Data as of 2026-08-30T08:39:29.467469+00:00.