ChenhongyiYang / SG-NMS

[ECCV 2020] Learning to Separate: Detecting Heavily-Occluded Objects in Urban Scenes

Learning to Separate: Detecting Heavily-Occluded Objects in Urban Scenes

This repository is the official implementation of our ECCV 2020 paper: Learning to Separate: Detecting Heavily-Occluded Objects in Urban Scenes

Requirements

Environment: You only need to install Tensorflow 1.13, and some common-used tools like opencv-python and numpy.

Data:
KITTI: Download the dataset from KITTI. To run the code, you need to first split the dataset into training and validation set by yourself.(Random partition is OK.)

CityPersons: Download the images from CityScapes and the annotations from here. You can use the official validation set to test your model.

Usage

1. Training

% train KITTI 
python bin/kitti/kitti_train.py --log True --gpu 0

% train CityPersons
python bin/citypersons/cp_trian.py --log True --gpu 0

2. Test

% test KITTI
python bin/kitti/kitti_test.py  --gpu 0

% test CityPersons
python bin/citypersons/cp_test.py --gpu 0

Citation

@inproceedings{yang2020separate,
  title={ Learning to Separate: Detecting Heavily-Occluded Objects in Urban Scenes},
  author={Yang, Chenhongyi and Ablavsky, Vitaly and Wang, Kaihong and Feng, Qi and Betke, Margrit},
  booktitle={ECCV},
  year={2020}
}