xjtAlgo / cascade-rcnn-fpn-faster_rcnn-pytorch1.0

This repo supports Faster R-CNN, FPN and Cascade Faster R-CNN based on pyTorch 1.0. Additionally deformable convolutional layer is also support!

Pytorch 1.0 Implementation of Faster R-CNN + FPN + Cascade Faster R-CNN

This repo supports Faster R-CNN, FPN and Cascade Faster R-CNN based on pyTorch 1.0. Additionally deformable convolutional layer is also support!

Train

Train Faster R-CNN on VOC dataset and pretrained resnet101.pth model

Before training, set the right directory to save and load the trained models. Change the arguments "save_dir" and "load_dir" in trainval_net.py and test_net.py to adapt to your environment.

python python train_faster_rcnn.py --dataset pascal_voc--net res101 --bs 1 --nw 1 --lr 0.01 --lr_decay_step 8 --cuda

Change dataset to "coco" or 'vg' if you want to train on COCO or Visual Genome.

Train FPN on VOC dataset and pretrained resnet101.pth model

Before training, set the right directory to save and load the trained models. Change the arguments "save_dir" and "load_dir" in trainval_net.py and test_net.py to adapt to your environment.

The learning rate should be lower than Faster RCNN

python python train_fpn.py --dataset pascal_voc--net res101 --bs 1 --nw 1 --lr 0.001 --lr_decay_step 8 --cuda

Change dataset to "coco" or 'vg' if you want to train on COCO or Visual Genome.

Train Cascade RCNN on VOC dataset and pretrained resnet101.pth model

Before training, set the right directory to save and load the trained models. Change the arguments "save_dir" and "load_dir" in trainval_net.py and test_net.py to adapt to your environment.

The learning rate should be lower than Faster RCNN

python python train_cascade_fpn.py --dataset pascal_voc--net res101 --bs 1 --nw 1 --lr 0.001 --lr_decay_step 8 --cuda

Change dataset to "coco" or 'vg' if you want to train on COCO or Visual Genome.

Test Faster R-CNN and generate json outputs

If you want to evlauate the detection performance of a pre-trained res101 model on pascal_voc test set, simply run

python json_test_faster_rcnn.py --dataset pascal_voc --net res101 \
                   --checksession $SESSION --checkepoch $EPOCH --checkpoint $CHECKPOINT \
                   --cuda

Specify the specific model session, chechepoch and checkpoint, e.g., SESSION=1, EPOCH=6, CHECKPOINT=416.

Test FPN and generate json outputs

If you want to evlauate the detection performance of a pre-trained res101 model on pascal_voc test set, simply run

python json_test_fpn.py --dataset pascal_voc --net res101 \
                   --checksession $SESSION --checkepoch $EPOCH --checkpoint $CHECKPOINT \
                   --cuda

Specify the specific model session, chechepoch and checkpoint, e.g., SESSION=1, EPOCH=6, CHECKPOINT=416.

Test Cascade R-CNN and generate json outputs

If you want to evlauate the detection performance of a pre-trained res101 model on pascal_voc test set, simply run

python json_test_cascade_fpn.py --dataset pascal_voc --net res101 \
                   --checksession $SESSION --checkepoch $EPOCH --checkpoint $CHECKPOINT \
                   --cuda

Specify the specific model session, chechepoch and checkpoint, e.g., SESSION=1, EPOCH=6, CHECKPOINT=416.

Citation

@article{jjfaster2rcnn,
    Author = {Jianwei Yang and Jiasen Lu and Dhruv Batra and Devi Parikh},
    Title = {A Faster Pytorch Implementation of Faster R-CNN},
    Journal = {https://github.com/jwyang/faster-rcnn.pytorch},
    Year = {2017}
}

@inproceedings{renNIPS15fasterrcnn,
    Author = {Shaoqing Ren and Kaiming He and Ross Girshick and Jian Sun},
    Title = {Faster {R-CNN}: Towards Real-Time Object Detection
             with Region Proposal Networks},
    Booktitle = {Advances in Neural Information Processing Systems ({NIPS})},
    Year = {2015}
}