Official PyTorch implementation of ObjectFormer for Image Manipulation Detection and Localization, CVPR 2022.
Please first download the imagenet-pretrained model from this Google Drive link.
Then for training, you could use:
python tools/run.py --cfg configs/objectformer_bs24_lr2.5e-4.yaml
while for evaluation, you could set TRAIN.ENABLE to False. For a better peformance on Pixel F1, you should adjust the TEST.THRES (0.5 by default) on each testing dataset.
Note that we only release the checkpoints trained on the publicly available dataset (CASIAV2 and IMD20). For a fair comparison with our method, please finetune it with your data.
If you find this repository helpful, please consider citing:
@inproceedings{wang2022objectformer,
title={Objectformer for image manipulation detection and localization},
author={Wang, Junke and Wu, Zuxuan and Chen, Jingjing and Han, Xintong and Shrivastava, Abhinav and Lim, Ser-Nam and Jiang, Yu-Gang},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
year={2022}
}