I find a conflict between this code and the relative paper!

[yhao-z]

• In this code, the CNNs in SLR-Net in model.py is defined as

      self.conv_1 = CONV_OP(n_f=16, ifactivate=True)
      self.conv_2 = CONV_OP(n_f=16, ifactivate=True)
      self.conv_3 = CONV_OP(n_f=16, ifactivate=False)
      self.conv_4 = CONV_OP(n_f=16, ifactivate=True)
      self.conv_5 = CONV_OP(n_f=16, ifactivate=True)
      self.conv_6 = CONV_OP(n_f=2, ifactivate=False)

• The channel of the 3d CNN is 16 that is different with 32 in your paper, and the conv_3 op should have a relu activation.
• However, if I change the code into

      self.conv_1 = CONV_OP(n_f=32, ifactivate=True)
      self.conv_2 = CONV_OP(n_f=32, ifactivate=True)
      self.conv_3 = CONV_OP(n_f=32, ifactivate=True)
      self.conv_4 = CONV_OP(n_f=32, ifactivate=True)
      self.conv_5 = CONV_OP(n_f=32, ifactivate=True)
      self.conv_6 = CONV_OP(n_f=2, ifactivate=False)

  to be consistent with the original paper, the total parameters of this SLR-Net is equal to 1140528 which is different with 441889 in the paper.

• Please confirm this conflict. It really matters.
  • I'm going to contrast your methods with mine, so it really matters to confirm that the results I reproduced are correct.

[Keziwen]

Hi Yinghao,

Thanks for your attention to our work! The parameter, 'niter', may should be set to 5. Then you can see if the total parameters is equal to 441889. ISTA-Net can be used as a baseline, as it differs from SLR-Net mainly in the LR Module. (5 * 4 = 441889 - 441869).

You can also communicate with me through the wechat for more detailed discussions. My wechat ID is: kzw_love_nw.

Best, Ziwen

[yhao-z]

Thanks very much！

[Angqi666]

Dear author,

We trying to reimplentment your method in our work with our data. I got some problem with the

 if acc == 8:
        mask = mat73.loadmat('/data1/ziwenke/SLRNet/mask_newdata/vista_' + mask_size + '_acc_8.mat')['mask']
  elif acc == 10:
      mask = mat73.loadmat('/data1/ziwenke/SLRNet/mask_newdata/vista_' + mask_size + '_acc_10.mat')['mask']
  elif acc == 12:
      mask = mat73.loadmat('/data1/ziwenke/SLRNet/mask_newdata/vista_' + mask_size + '_acc_12.mat')['mask']

what is the shape of the mask here?

Best, Angqi