Count- and Similarity-aware RCNN for Pedestrian Detection

Pedestrian detection framework as detailed in [paper], accepted to ECCV 2020.

Recent pedestrian detection methods generally rely on additional supervision, such as visible bounding-box annotations, to handle heavy occlusions. We propose an approach that leverages pedestrian count and proposal similarity information within a two-stage pedestrian detection framework. Both pedestrian count and proposal similarity are derived from standard full-body annotations commonly used to train pedestrian detectors. We introduce a count-weighted detection loss function that assigns higher weights to the detection errors occurring at highly overlapping pedestrians. The proposed loss function is utilized at both stages of the two-stage detector. We further introduce a count-and-similarity branch within the two-stage detection framework, which predicts pedestrian count as well as proposal similarity. Lastly, we introduce a count and similarity-aware NMS strategy to identify distinct proposals. Our approach requires neither part information nor visible bounding-box annotations.

pip install -r requirements.txt
python setup.py develop

It is recommended to symlink the dataset root to $CaSe/data. You can download CityScapes Datasets.Put it in data folder.

CaSe
├── data
│   ├── cityscapes
│   │   ├── leftImg8bit
│   │   │   ├── train
│   │   │   ├── val

./tools/dist_test.sh ${CONFIG_FILE} ${CHECKPOINT_FILE} ${GPU_NUM}

• CONFIG_FILE about CaSe is in configs.

We provide a demo script to test a single image.

python tools/demo.py ${IMAGE_FILE} ${CONFIG_FILE} ${CHECKPOINT_FILE}

If the project helps your research, please cite this paper.

@InProceedings{Xie_CaSe_ECCV_2020,
  author =       {Jin Xie and Hisham Cholakkal and Rao Muhammad Anwer and Fahad Shahbaz Khan and Yanwei Pang and Ling Shao and Mubarak Shah},
  title =        {Count- and Similarity-aware RCNN for Pedestrian Detection},
  booktitle =    {The European Conference on Computer Vision (ECCV)},
  year =         {2020}
}

Many thanks to the open source codes, i.e., mmdetection.