Progressive Refinement Network for Occluded Pedestrian Detection

Keras implementation of PRNet accepted in ECCV 2020.

Dependencies

• python 2.7
• numpy 1.12.0
• Tensorflow 1.x
• keras 2.0.6
• OpenCV

Get Start

1. Get the code.

git clone https://github.com/sxlpris/PRNet.git

2. Install the requirements.

  pip install -r requirements.txt

3. Download the dataset CityPersons to folder '$PRNet/data/cityperson/'.

4. Download the initialized models ResNet-50 to folder '$PRNet/data/models/'.

5. Train.

run *train_prnet.py*

5. Test.

run *test_prnet.py*

Model

To help reproduce the results in our paper, we provide our model PRNet_city.hdf5 (password:imiq) trained on CityPersons.

Ackownledgment

The code of our work is based on the pipeline of "https://github.com/liuwei16/ALFNet". Thanks for the code and training protocol.