An implementation of Cascade R-CNN: Delving into High Quality Object Detection. I only trained and tested on pascal voc dataset. The source code is here which implemented by caffe and also evalated on pascal voc.

As we all know, the cascade structure is designed for R-CNN structure, so i just used the cascade structure based on DetNet to train and test on pascal voc dataset (DetNet is not only faster than fpn-resnet101, but also better than fpn-resnet101).

Based on DetNet_Pytorch, i mainly changed the forward function in fpn.py. It‘s just a naive implementation, so its speed is not fast.

2019/01/01:

• Fix bugs in demo, now you can run demo.py file. Note the default demo.py merely support pascal_voc categories. You need to change the pascal_classes in demo.py to adapt your own dataset. If you want to know more details, please see the usage part.
• upload the pretrained DetNet59-Cascade which in below table.

I benchmark this code thoroughly on pascal voc2007 and 07+12. Below are the results:

1). PASCAL VOC 2007 (Train/Test: 07trainval/07test, scale=600, ROI Align)

2). PASCAL VOC 07+12 (Train/Test: 07+12trainval/07test, scale=600, ROI Align)

First of all, clone the code

git clone https://github.com/guoruoqian/cascade-rcnn_Pytorch.git

Then, create a folder:

cd cascade-rcnn_Pytorch && mkdir data

• Python 2.7 or 3.6
• Pytorch 0.2.0 or higher（not support pytorch version >=0.4.0）
• CUDA 8.0 or higher
• tensorboardX

• VOC2007: Please follow the instructions in py-faster-rcnn to prepare VOC datasets. Actually, you can refer to any others. After downloading the data, creat softlinks in the folder data/.
• VOC 07 + 12: Please follow the instructions in YuwenXiong/py-R-FCN . I think this instruction is more helpful to prepare VOC datasets.

 You can download the detnet59 model which i trained on ImageNet from:

• detnet59: dropbox，baiduyun，google drive

Download it and put it into the data/pretrained_model/.

As pointed out by ruotianluo/pytorch-faster-rcnn, choose the right -arch in make.sh file, to compile the cuda code:

Install all the python dependencies using pip:

pip install -r requirements.txt

Compile the cuda dependencies using following simple commands:

cd lib
sh make.sh

It will compile all the modules you need, including NMS, ROI_Pooing, ROI_Align and ROI_Crop. The default version is compiled with Python 2.7, please compile by yourself if you are using a different python version.

If you want to use cascade structure, you must set --cascade and --cag in the below script. cag determine whether perform class_agnostic bbox regression.

train voc2007 use cascade structure:

CUDA_VISIBLE_DEVICES=3 python3 trainval_net.py exp_name --dataset pascal_voc --net detnet59 --bs 2 --nw 4 --lr 1e-3 --epochs 12 --save_dir weights --cuda --use_tfboard True --cag --cascade

test voc2007:

CUDA_VISIBLE_DEVICES=3 python3 test_net.py exp_name --dataset pascal_voc --net detnet59 --checksession 1 --checkepoch 7 --checkpoint 5010 --cuda --load_dir weights --cag --cascade

Before training voc07+12, you must set ASPECT_CROPPING in detnet59.yml False, or you will encounter some error during the training.

train voc07+12:

CUDA_VISIBLE_DEVICES=3 python3 trainval_net.py exp_name2 --dataset pascal_voc_0712 --net detnet59 --bs 1 --nw 4 --lr 1e-3 --epochs 12 --save_dir weights --cuda --use_tfboard True --cag --cascade

run demo.py :

Before run demo, you must make dictionary 'demo_images' and put images (VOC images) in it. You can download the pretrained model  listed in above tables.

CUDA_VISIBLE_DEVICES=3 python3 demo.py exp_name2 --dataset pascal_voc_0712 --net detnet59 --checksession 1 --checkepoch 8 --checkpoint 33101 --cuda --load_dir weights --cag --image_dir demo_images --cascade --result_dir vis_cascade