xingyizhou/CenterNet2

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1219 stars · 189 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

CenterNet2 is a research implementation of probabilistic two-stage object detection built on detectron2, where a class-agnostic one-stage CenterNet proposes objects and a second stage classifies them. It includes pre-trained COCO, LVIS, and Objects365 models reaching up to 56.4 mAP on COCO test-dev.

Use cases

• train a two-stage object detector on COCO
• run object detection inference on images with pre-trained models
• improve detection accuracy over one-stage CenterNet
• benchmark object detectors on LVIS or Objects365
• use CenterNet as a proposal network for detection

When to choose

• you need state-of-the-art two-stage detection accuracy with a detectron2 workflow
• you want pre-trained models for COCO, LVIS, or Objects365
• you are reproducing the 'Probabilistic two-stage detection' paper

When to avoid

• you need a lightweight production detector with active maintenance
• you are not already using the detectron2 ecosystem
• you need Windows or CPU-only deployment support

Facets

library · maturity maintenance

computer-vision machine-learning deep-learning computer-vision image-processing deep-learning python object-detection centernet detectron2 coco two-stage-detector research-code gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="xingyizhou/CenterNet2")