weizequan / NIvsCI

Improving the Generalization of Colorized Image Detection with Enhanced Training of CNN

NIvsCI

Prerequisites

• Python 2
• NVIDIA GPU + CUDA cuDNN
• PyTorch 0.3.1

Run

• Normal training: python normal_training.py
• Construct negative samples in the offline manner: python constructNegativeSamples.py
• Enhanced training: python enhanced_training.py
• Test the results: python test.py

Citation

If you find this useful for your research, please use the following.

@INPROCEEDINGS{quan2019enhance, 
  author={W. {Quan} and K. {Wang} and D. {Yan} and D. {Pellerin} and X. {Zhang}}, 
  booktitle={Proceedings of the International Symposium on Image and Signal Processing and Analysis}, 
  title={Improving the Generalization of Colorized Image Detection with Enhanced Training of CNN}, 
  year={2019}, 
  pages={246-252}
}

Acknowledgments

This code borrows from pytorch-CycleGAN-and-pix2pix.