xingyizhou/CenterNet

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

7573 stars · 1934 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

CenterNet is a PyTorch implementation of the 'Objects as Points' detector, which models objects as single center points detected via keypoint estimation, regressing size, 3D location, orientation, and pose. It supports object detection on COCO, 3D bounding box detection on KITTI, and multi-person pose estimation in a single end-to-end network without NMS.

Use cases

• detect objects in images in real time
• estimate 3D bounding boxes from monocular images
• run multi-person human pose estimation
• train a fast anchor-free object detector on custom datasets
• benchmark speed-accuracy trade-offs on COCO

When to choose

• you need a simple, fast, anchor-free detector without NMS post-processing
• you want one framework covering 2D detection, 3D detection, and pose estimation
• you need real-time inference on a GPU

When to avoid

• you need actively maintained code with recent framework compatibility
• you require state-of-the-art transformer-based detectors like DETR variants
• you need Lidar-based 3D detection (see CenterPoint instead)

Facets

library · maturity maintenance

computer-vision image-processing machine-learning deep-learning computer-vision deep-learning python object-detection centernet keypoint-estimation 3d-detection pose-estimation pytorch research-code linux gpu

Member repositories

For agents

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