This repository contains all of our code for DSAI | K64 | HUST | Computer Vision project.
To download the dataset, run this code:
wget http://host.robots.ox.ac.uk/pascal/VOC/voc2012/VOCtrainval_11-May-2012.tar
tar -xf VOCtrainval_11-May-2012.tar
mkdir PascalVOC12
mv VOCdevkit/VOC2012/* PascalVOC12
cd PascalVOC12
wget http://cs.jhu.edu/~cxliu/data/SegmentationClassAug.zip
wget http://cs.jhu.edu/~cxliu/data/SegmentationClassAug_Visualization.zip
wget http://cs.jhu.edu/~cxliu/data/list.zip
unzip SegmentationClassAug.zip
unzip SegmentationClassAug_Visualization.zip
unzip list.zip
mv list splits
Refactor the dataroot as follow:
data_root/
--- VOC2012/
--- Annotations/
--- ImageSet/
--- JPEGImages/
--- SegmentationClassAug/
--- saliency_map/
--- ADEChallengeData2016
--- annotations
--- training
--- validation
--- images
--- training
--- validation
Download SegmentationClassAug and saliency_map
Check the demo for a intuitive guideline. In particular, run the command:
python main.py --data_root ${DATA_ROOT} --approach ${APPROACH} --model deeplabv3_resnet101 --gpu_id 0,1 --crop_val --lr ${LR} --batch_size ${BATCH} --train_epoch ${EPOCH} --loss_type ${LOSS} --dataset ${DATASET} --task ${TASK} --overlap --lr_policy poly --pseudo --pseudo_thresh ${THRESH} --freeze --bn_freeze --unknown --w_transfer --amp --mem_size ${MEMORY}
To perform the original version (network frozen version), run this command:
$ python main.py --data_root [your data root] --approach css --model deeplabv3_resnet101 --gpu_id 0,1 --crop_val --lr 0.01 --batch_size 16 --train_epoch 50 --loss_type bce_loss --dataset voc --task 5-5 --overlap --lr_policy poly --pseudo --pseudo_thresh 0.7 --freeze --bn_freeze --unknown --w_transfer --amp --mem_size 100 --val_interval 100
To perform the variant version (network masking using hard attention), run this command:
$ python main.py --data_root [your data root] --approach hat_ss --model deeplabv3_resnet101 --overlap --gpu_id 0,1 --crop_val --lr 0.01 --bb_lr 0.0001 --batch_size 16 --train_epoch 50 --loss_type bce_loss --dataset voc --task 2-1 --lr_policy poly --pseudo --pseudo_thresh 0.7 --freeze --bn_freeze --unknown --w_transfer --amp --mem_size 100 --val_interval 100 --lamb 1
Our implementation is based on clovaai/SSUL, k-gyuhak/CLOM.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent