Home

Awesome

Simple multi-dataset detection

An object detector trained on multiple large-scale datasets with a unified label space; Winning solution of ECCV 2020 Robust Vision Challenges.

<p align="center"> <img src='docs/unidet_teaser.jpg' align="center" height="170px"> </p>

Simple multi-dataset detection,
Xingyi Zhou, Vladlen Koltun, Philipp Krähenbühl,
CVPR 2022 (arXiv 2102.13086)

Contact: zhouxy@cs.utexas.edu. Any questions or discussions are welcomed!

Features at a glance

Main results

COCO test-challengeOpenImages public testMapillary testObjects365 val
52.960.625.333.7

Results are obtained using a Cascade-RCNN with ResNeSt200 trained in an 8x schedule.

COCOObjects365OpenImagesmean.
Unified45.424.466.045.3
Dataset-specific models42.524.965.744.4

Results are obtained using a Cascade-RCNN with Res50 trained in an 8x schedule.

VOCVIPERCityScapesScanNetWildDashCrowdHumanKITTImean
Unified82.921.352.629.834.770.739.947.3
Oracle models80.331.854.644.7-80.0--

Results are obtained using a Cascade-RCNN with Res50 trained in an 8x schedule.

More models can be found in our MODEL ZOO.

Installation

Our project is developed on detectron2. Please follow the official detectron2 installation.

Demo

We use the same inference API as detectorn2. To run inference on an image folder using our pretrained model, run

python demo.py --config-file configs/Unified_learned_OCIM_R50_6x+2x.yaml --input images/*.jpg --opts MODEL.WEIGHTS models/Unified_learned_OCIM_R50_6x+2x.pth

If setup correctly, the output should look like:

<p align="center"> <img src='docs/example_output2.jpg' align="center" height="460px"> </p>

*The sample image is from WildDash dataset.

Note that the model predicts all labels in its label hierarchy tree (for example, both vehicle and car for a car), following the protocol in OpenImages.

Benchmark evaluation and training

After installation, follow the instructions in DATASETS.md to setup the (many) datasets. Then check REPRODUCE.md to reproduce the results in the paper.

License

Our code is under Apache 2.0 license.

Citation

If you find this project useful for your research, please use the following BibTeX entry.

@inproceedings{zhou2021simple,
  title={Simple multi-dataset detection},
  author={Zhou, Xingyi and Koltun, Vladlen and Kr{\"a}henb{\"u}hl, Philipp},
  booktitle={CVPR},
  year={2022}
}