[CVPR 2022] Contextual Outpainting With Object-Level Contrastive Learning

Jiacheng Li, Chang Chen, and Zhiwei Xiong

CVPR Paper | CVPR Paper Supp. | Project Page | Intro Video | Demo Video

Folder structure

Please download the coco val2017 stuffmask and name.lst from Release and organize them as follows.

# test dataset
datasets/coco/val2017/images    # please download from cocodataset.org
                     /stuffmask
                     /name.list

Next, download pretrained models from Release and put them into the default checkpoints folder.

# pretrained models
checkpoints/$con_name/latest_*.pth
           /$syn_name/latest_*.pth

Finally, run the following command to start testing.

# test.sh
python test_2stage.py \
    --data_root datasets/coco/val2017 \     # dataset root
    --test_name ctogan_test \               # save path: results/test_name
    --con_name bc_context \                 # semantic reasoning model path
    --syn_name condspade \                  # content generation model path
    --e_nsampling 3                         # number of sample
    

Citation

@InProceedings{Li_2022_CVPR,
    author    = {Li, Jiacheng and Chen, Chang and Xiong, Zhiwei},
    title     = {Contextual Outpainting With Object-Level Contrastive Learning},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2022},
    pages     = {11451-11460}
}