- Install Pytorch>=1.0.0 in python3.
- Clone this repository.
git clone https://github.com/zhangminwen/Center-and-Scale-Prediction-CSP-Pytorch.git
- Download our pre-trained model and save in $CSP_ROOT/weights. model:epoch320 password:s06j
- Test the model by using gpu.
cd $CSP_ROOT
python ./test.py
- Test the model by using cpu.
python ./test.py --cpu
The "result.jpg" is detection result.
-
Prepare the dataset in $CSP_ROOT/city.
The pedestrian annotations (xmin, ymin, xmax, ymax) in each image are stored in the txt document which name correspond to the image name and saved in $CSP_ROOT/city/Annotations.
Images are saved in $CSP_ROOT/city/Images.
Splits of training, testing, valudation are saved in $CSP_ROOT/city/ImageSets.
-
Download pretrained ResNet50 model and save in $CSP_ROOT/model.
-
Train
python ./train.py
@inproceedings{liu2018high,
title={High-level Semantic Feature Detection: A New Perspective for Pedestrian Detection},
author={Wei Liu, Shengcai Liao, Weiqiang Ren, Weidong Hu, Yinan Yu},
booktitle={IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
year={2019}
}