https://github.com/xingyizhou/CenterNet2
Two-stage CenterNet
https://github.com/xingyizhou/CenterNet2
coco object-detection
Last synced: about 1 month ago
JSON representation
Two-stage CenterNet
- Host: GitHub
- URL: https://github.com/xingyizhou/CenterNet2
- Owner: xingyizhou
- License: apache-2.0
- Created: 2021-03-15T00:27:22.000Z (about 4 years ago)
- Default Branch: master
- Last Pushed: 2022-11-20T23:07:54.000Z (over 2 years ago)
- Last Synced: 2024-10-29T17:49:47.554Z (6 months ago)
- Topics: coco, object-detection
- Language: Python
- Homepage:
- Size: 4.88 MB
- Stars: 1,206
- Watchers: 20
- Forks: 188
- Open Issues: 58
-
Metadata Files:
- Readme: README.md
- Contributing: .github/CONTRIBUTING.md
- License: LICENSE
- Code of conduct: .github/CODE_OF_CONDUCT.md
Awesome Lists containing this project
- awesome-anchor-free-object-detection - CenterNet2 - stage detection". (**[arXiv 2021](https://arxiv.org/abs/2103.07461)**) (Frameworks)
README
# Probabilistic two-stage detection
Two-stage object detectors that use class-agnostic one-stage detectors as the proposal network.
> [**Probabilistic two-stage detection**](http://arxiv.org/abs/2103.07461),
> Xingyi Zhou, Vladlen Koltun, Philipp Krähenbühl,
> *arXiv technical report ([arXiv 2103.07461](http://arxiv.org/abs/2103.07461))*
Contact: [[email protected]](mailto:[email protected]). Any questions or discussions are welcomed!
## Summary
- Two-stage CenterNet: First stage estimates object probabilities, second stage conditionally classifies objects.
- Resulting detector is faster and more accurate than both traditional two-stage detectors (fewer proposals required), and one-stage detectors (lighter first stage head).
- Our best model achieves 56.4 mAP on COCO test-dev.
- This repo also includes a detectron2-based CenterNet implementation with better accuracy (42.5 mAP at 70FPS) and a new FPN version of CenterNet (40.2 mAP with Res50_1x).
## Main results
All models are trained with multi-scale training, and tested with a single scale. The FPS is tested on a Titan RTX GPU.
More models and details can be found in the [MODEL_ZOO](docs/MODEL_ZOO.md).
#### COCO
| Model | COCO val mAP | FPS |
|-------------------------------------------|---------------|-------|
| CenterNet-S4_DLA_8x | 42.5 | 71 |
| CenterNet2_R50_1x | 42.9 | 24 |
| CenterNet2_X101-DCN_2x | 49.9 | 8 |
| CenterNet2_R2-101-DCN-BiFPN_4x+4x_1560_ST | 56.1 | 5 |
| CenterNet2_DLA-BiFPN-P5_24x_ST | 49.2 | 38 |
#### LVIS
| Model | val mAP box |
| ------------------------- | ----------- |
| CenterNet2_R50_1x | 26.5 |
| CenterNet2_FedLoss_R50_1x | 28.3 |
#### Objects365
| Model | val mAP |
|-------------------------------------------|----------|
| CenterNet2_R50_1x | 22.6 |
## Installation
Our project is developed on [detectron2](https://github.com/facebookresearch/detectron2). Please follow the official detectron2 [installation](https://github.com/facebookresearch/detectron2/blob/master/INSTALL.md).
We use the default detectron2 demo script. To run inference on an image folder using our pre-trained model, run
~~~
python demo.py --config-file configs/CenterNet2_R50_1x.yaml --input path/to/image/ --opts MODEL.WEIGHTS models/CenterNet2_R50_1x.pth
~~~
## Benchmark evaluation and training
Please check detectron2 [GETTING_STARTED.md](https://github.com/facebookresearch/detectron2/blob/master/GETTING_STARTED.md) for running evaluation and training. Our config files are under `configs` and the pre-trained models are in the [MODEL_ZOO](docs/MODEL_ZOO.md).
## License
Our code is under [Apache 2.0 license](LICENSE). `centernet/modeling/backbone/bifpn_fcos.py` are from [AdelaiDet](https://github.com/aim-uofa/AdelaiDet), which follows the original [non-commercial license](https://github.com/aim-uofa/AdelaiDet/blob/master/LICENSE).
## Citation
If you find this project useful for your research, please use the following BibTeX entry.
@inproceedings{zhou2021probablistic,
title={Probabilistic two-stage detection},
author={Zhou, Xingyi and Koltun, Vladlen and Kr{\"a}henb{\"u}hl, Philipp},
booktitle={arXiv preprint arXiv:2103.07461},
year={2021}
}