Are you sure you want to delete this access key?
Legend |
---|
DVC Managed File |
Git Managed File |
Metric |
Stage File |
External File |
Legend |
---|
DVC Managed File |
Git Managed File |
Metric |
Stage File |
External File |
conda create -n zsi python=3.7 -y
conda activate zsi
conda install pytorch=1.1.0 torchvision=0.3.0 cudatoolkit=10.0 -c pytorch
pip install cython && pip --no-cache-dir install -r requirements.txt
python setup.py develop
Download the train and test annotations files for zsi from annotations, put all json label file to
data/coco/annotations/
Download MSCOCO-2014 dataset and unzip the images it to path:
data/coco/train2014/
data/coco/val2014/
Training:
48/17 split:
chmod +x tools/dist_train.sh
./tools/dist_train.sh configs/zsi/train/zero-shot-mask-rcnn-BARPN-bbox_mask_sync_bg_decoder.py 4
65/15 split:
chmod +x tools/dist_train.sh
./tools/dist_train.sh configs/zsi/train/zero-shot-mask-rcnn-BARPN-bbox_mask_sync_bg_65_15_decoder_notanh.py 4
Inference & Evaluate:
ZSI task:
download 48/17 ZSI model, put it in checkpoints/ZSI_48_17.pth
inference:
chmod +x tools/dist_test.sh
./tools/dist_test.sh configs/zsi/48_17/test/zsi/zero-shot-mask-rcnn-BARPN-bbox_mask_sync_bg_decoder.py checkpoints/ZSI_48_17.pth 4 --json_out results/zsi_48_17.json
our results zsi_48_17.bbox.json and zsi_48_17.segm.json can also downloaded from zsi_48_17_reults.
evaluate:
python tools/zsi_coco_eval.py results/zsi_48_17.bbox.json --ann data/coco/annotations/instances_val2014_unseen_48_17.json
python tools/zsi_coco_eval.py results/zsi_48_17.segm.json --ann data/coco/annotations/instances_val2014_unseen_48_17.json --types segm
download 65/15 ZSI model, put it in checkpoints/ZSI_65_15.pth
inference:
chmod +x tools/dist_test.sh
./toools/dist_test.sh configs/zsi/65_15/test/zsi/zero-shot-mask-rcnn-BARPN-bbox_mask_sync_bg_65_15_decoder_notanh.py checkpoints/ZSI_65_15.pth 4 --json_out results/zsi_65_15.json
our results zsi_65_15.bbox.json and zsi_65_15.segm.json can also downloaded from zsi_65_15_reults.
evaluate:
python tools/zsi_coco_eval.py results/zsi_65_15.bbox.json --ann data/coco/annotations/instances_val2014_unseen_65_15.json
python tools/zsi_coco_eval.py results/zsi_65_15.segm.json --ann data/coco/annotations/instances_val2014_unseen_65_15.json --types segm
GZSI task:
chmod +x tools/dist_test.sh
./tools/dist_test.sh configs/zsi/48_17/test/gzsi/zero-shot-mask-rcnn-BARPN-bbox_mask_sync_bg_decoder_gzsi.py checkpoints/ZSI_48_17.pth 4 --json_out results/gzsi_48_17.json
python tools/gzsi_coco_eval.py results/gzsi_48_17.bbox.json --ann data/coco/annotations/instances_val2014_gzsi_48_17.json --gzsi --num-seen-classes 48
python tools/gzsi_coco_eval.py results/gzsi_48_17.segm.json --ann data/coco/annotations/instances_val2014_gzsi_48_17.json --gzsi --num-seen-classes 48 --types segm
chmod +x tools/dist_test.sh
./tools/dist_test.sh configs/zsi/65_15/test/gzsi/zero-shot-mask-rcnn-BARPN-bbox_mask_sync_bg_65_15_decoder_notanh_gzsi.py checkpoints/ZSI_65_15.pth 4 --json_out results/gzsi_65_15.json
python tools/gzsi_coco_eval.py results/gzsi_65_15.bbox.json --ann data/coco/annotations/instances_val2014_gzsi_65_15.json --gzsd --num-seen-classes 65
python tools/gzsi_coco_eval.py results/gzsi_65_15.segm.json --ann data/coco/annotations/instances_val2014_gzsi_65_15.json --gzsd --num-seen-classes 65 --types segm
ZSI is released under MIT License.
If you use ZSI in your research or wish to refer to the baseline results published here, please use the following BibTeX entries:
@InProceedings{Zheng_2021_CVPR,
author = {Zheng, Ye and Wu, Jiahong and Qin, Yongqiang and Zhang, Faen and Cui, Li},
title = {Zero-Shot Instance Segmentation},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2021},
pages = {2593-2602}
}
Press p or to see the previous file or, n or to see the next file
Are you sure you want to delete this access key?
Are you sure you want to delete this access key?
Are you sure you want to delete this access key?
Are you sure you want to delete this access key?