Unofficially Pytorch implementation of High-level Semantic Feature Detection: A New Perspective for Pedestrian Detection

This code is only for CityPersons dataset, and only for center-position+height regression+offset regression model.

This repo will no longer be updated. After all the experiments are done, I will open a new repo. Currently MR of CityPerson has reach 10.2%, time consuming 0.16 sec per frame(4K)

On Cityperson validation set

11.71 MR CSPNet-26.pth (NEW !)

12.56 MR CSPNet-89.pth

Python, pytorch and other related libaries

GPU is needed

Compile lib

cd util
make all

Prepare CityPersons dataset as the original codes doing

• For citypersons, we use the training set (2975 images) for training and test on the validation set (500 images), we assume that images and annotations are stored in ./data/citypersons, and the directory structure is

*DATA_PATH
	*annotations
		*anno_train.mat
		*anno_val.mat
	*images
		*train
		*val

Training & val

python trainval_torchstyle.py
python trainval_caffestyle.py

NOTE

using caffe style, you need to download additional pre-trained weight.