A Faster Pytorch Implementation for Caltech Pedestrian dataset

This implement is based on jwyang/faster-rcnn.pytorch. I make some changes to adapt the pedestrian detection. The Pytorch version is 4.0.

References

• jwyang/faster-rcnn.pytorch

• ruotianluo/pytorch-resnet

• GBJim/mscnn

• ChaoPei/faster-rcnn-pedestrian-detection

• CasiaFan/Dataset_to_VOC_converter

Tutorial

• use CasiaFan/Dataset_to_VOC_converter to create training data

• build modules

cd lib
sh make.sh

• train_single_gpu.sh for training on single gpu

sh train_single_gpu.sh 0

• train_mgpu.sh for trainning on multi-gpus

sh train_mgpus.sh

• predict single image

sh predict.sh

• create caltech eval result

sh caltech_dataset_detect.sh