Duankaiwen/CenterNet
CenterNet is a one-stage object detector that identifies objects by detecting center points and keypoint triplets.
Not currently ranked — collecting fresh signals.
star history
CenterNet is a deep learning object detection approach that builds upon keypoint-based detection but adds center keypoint prediction to address incorrect bounding boxes. The model uses a stacked hourglass backbone to predict center points, corners, and corner offsets simultaneously. On the MS-COCO dataset, it achieves 47.0% AP, surpassing all known one-stage detectors and approaching two-stage detector performance.
Frequently asked
- What is Duankaiwen/CenterNet?
- CenterNet is a one-stage object detector that identifies objects by detecting center points and keypoint triplets.
- Is CenterNet open source?
- Yes — Duankaiwen/CenterNet is open source, released under the MIT license.
- What language is CenterNet written in?
- Duankaiwen/CenterNet is primarily written in Python.
- How popular is CenterNet?
- Duankaiwen/CenterNet has 1.9k stars on GitHub.
- Where can I find CenterNet?
- Duankaiwen/CenterNet is on GitHub at https://github.com/Duankaiwen/CenterNet.