search

rmpku / CIN

Paper 'Towards Blind Watermarking: Combining Invertible and Non-invertible Mechanisms' in ACM Multimedia '22.
49 stars 3 forks source link

Confusion about parts of the code #4

Closed kerryhhh closed 1 year ago

kerryhhh commented 1 year ago

Thank you for sharing your code! But I'm still confused about parts of the code.

  1. I noticed that in the CIN code, there doesn't seem to be a training process for the NSM
  2. The JPEG compression seems to be slightly different from that in the paper I hope you can answer the above questions, thank you very much!
rmpku commented 1 year ago

Thank you for sharing your code! But I'm still confused about parts of the code.

  1. I noticed that in the CIN code, there doesn't seem to be a training process for the NSM
  2. The JPEG compression seems to be slightly different from that in the paper I hope you can answer the above questions, thank you very much!

1..The structure of NSM is described in the supplementary material (page 11 of the article), and it is easy to training to converge. 2..The code of JPEG Compression is an optimized version, and there are some inconsistencies in the article, thanks for your correction. In the formula on page 5 of the article, 'JpegMask' corresponds to 'Jpeg' in the code, and 'RealJpeg' corresponds to 'JpegTest' in the code (CIN.py).