# tryolabs/norfair

Lightweight Python library for adding real-time multi-object tracking to any detector.

Repository: https://github.com/tryolabs/norfair
Canonical: https://ross.abutalabs.com/products/norfair
Homepage: https://tryolabs.github.io/norfair/
Language: Python
License: BSD-3-Clause
License Family: permissive
Topics: tracking, tracking-algorithm, kalman-filter, object-tracking, video-tracking, video-inference-loop, deepsort, object-detection, pose-estimation, python, re-id, re-identification, multi-object-tracking
Last push: 2025-04-30T17:21:08+00:00

## Health v2 (maintenance only)
Score: 31/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 19, release rhythm 8, longevity 100
- inputs: {"age_days": 2254, "days_push": 490, "days_rel": 490, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2676, forks 275 (observed 2026-08-28T04:07:09.752893+00:00)

## What it is
Norfair is a lightweight, customizable Python library for real-time multi-object tracking that works with any detector outputting (x, y) coordinates. It supports moving cameras, re-identification with appearance embeddings, and n-dimensional tracking, and can be dropped into existing video pipelines or used to build inference loops from scratch.

## Use cases
- track players in sports video with a moving camera
- add multi-object tracking to my YOLO detector
- track pose keypoints across video frames
- count and follow objects in real-time video streams
- re-identify people after occlusion using appearance embeddings
- track objects in 3D point detections
- build a video inference loop with detection and tracking

## When to choose
- you need fast, real-time tracking on top of an existing detector
- you want a minimal, modular tracking library that fits into custom pipelines
- you need support for moving cameras or re-identification
- you track keypoints or arbitrary coordinate detections, not just bounding boxes

## When to avoid
- you need a full out-of-the-box video analytics application with UI and dashboards
- you require heavy-duty trackers with built-in state-of-the-art transformer-based association
- your project is not in Python

## Facets
- artifact type: library
- maturity: stable
- function: computer-vision, video-processing, machine-learning
- domain: computer-vision, machine-learning
- platform: python, cross-platform
- tags: multi-object-tracking, kalman-filter, object-detection, pose-estimation, re-identification, deepsort, real-time, video

## Member repositories
- tryolabs/norfair (main) score 31

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:09.752893+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 2026-08-30T02:18:02.114187+00:00, confidence not recorded.
  - readme: https://github.com/tryolabs/norfair (fetched 2026-08-28T04:07:09.752893+00:00, sha ef7a95b387b0)
  - homepage: https://tryolabs.github.io/norfair/ (fetched 2026-08-29T10:00:30.767465+00:00, sha 36c6c3c2e4f9)
  - registry_pypi: https://pypi.org/pypi/norfair/json (fetched 2026-08-29T10:00:30.769842+00:00, sha 87bbe92011e7)
- Data as of 2026-08-30T08:39:29.467469+00:00.