PHNet: Parasite-Host Network for Video Crowd Counting

Training

Pretreatment

Download Venice Dataset as example, unzip and move Venice dataset into ./PHNet-pytorch/dataset/Venice
Perform pretreatments in order(density map, ROI and 3D dataset generation)

Train

From initial

python model/train.py /dataset/Venice/train_data.json /dataset/Venice/test_data.json 1 --gpu 0,1 -bs 8

From pretrained model

cd model && wget https://github.com/LeeJAJA/PHNet-pytorch/releases/download/1.0/PHnet_checkpoint.pth

python model/train.py /dataset/Venice/train_data.json /dataset/Venice/test_data.json 1 --gpu 0,1 -bs 8 --pre /model/PHnet_checkpoint.pth

Authors & Contributors

PHNet is authored by Jiajie Li, Shiqiao Meng, Weiwei Guo, Lai Ye and Jinfeng Jiang from Tongji University. Jiajie Li is the corresponding author. The code is developed by Jiajie Li and Shiqiao Meng. Currently, it is being maintained by Jiajie Li.