fenghetan9 / multi-level-global-context-cross-consistency Goto Github PK

Hi, I would like to know how your dataset is stored, I read your two papers and found that the dataset storage format is from the last CMU, I would like to know how you store the unlabeled images generated by diffusion modeling into the dataset and for semi-supervised training dataset how it should be stored.

Hi, I generated the BUSI data set according to the guidance of your paper, but the quality of my pictures is not as high as that of your paper. Could you please provide the parameters of your generative diffusion model? If you can provide me Would be very grateful!

hi, thanks for you job! Here is a question about dataloader about labeled dataset and unlabeled dataset. Here a set a semi percentage about 0.5, the batch_size is 8 and labeled_bs is 4, and I set the idx around [0,1903] are unlabeled and else are labeled, but when I call the dataloader, the grouper function is normal, the first four are labeled and the last four are unlabeled, but during the training stage, i found there goes a mixture in labeled idx and unlabeled idx, for example, there is an error about the unlabled mask cannot found in the masked set, is there anything error in the getitem function because it returns both image and label. Thanks a lot!

Hi
Thanks for your work on this repo!

What would you recommend to use as the pre-trained model?
Have you used and seen an improvement if finetuning a pre-trained model?

Thanks in advance

Is there a code for the inference ？especially for the experiment analyzed the average inference time and GFLOPs, Thank you

Hi

Does this code include the generate stage (using the Diffusion) or only the semi supervised learning stage.

Hello, I have a problem generating images, and when I use them, I find that the generated pictures are randomly generated and repetitive, I would like to ask you how to set up the complete generation of pictures corresponding to the input

Hi, We ran the code for 295 epochs. Below is the log after the run of the code. Please help us if we are missing something

epoch [294/295] train_loss 0.2000 supervised_loss 0.1954 consistency_loss 0.0012 train_iou: 0.9596 - val_loss 0.5416 - val_iou 0.6689 - val_SE 0.5690 - val_PC 0.6468 - val_F1 0.5644 - val_ACC 0.7565

We have done this modification for learning rate, as we were encountering the "RuntimeError: For non-complex input tensors, argument alpha must not be a complex number.". BAsed on the link provided by you in other issue

def adjust_learning_rate(optimizer, i_iter, len_loader, max_epoch, power, args):
lr = lr_poly(args.base_lr, i_iter, max_epoch*len_loader, power)
optimizer.param_groups[0]['lr'] = lr
if len(optimizer.param_groups) > 1:
optimizer.param_groups[1]['lr'] = lr * 10
return lr

lr_ = adjust_learning_rate(optimizer, iter_num, len(trainloader), max_epoch, 0.9, args)

assert len(self.secondary_indices) >= self.secondary_batch_size > 0
AssertionError

Hello, I have successfully trained your code. May I ask if there is any code that divides parts? I couldn't find it in it. I would greatly appreciate it if you could submit the code for the testing section

Hi
Thanks for sharing!

I found the test model(UNet) you used contians ConvMixer layer in your paper. But in the Evaluation Metrics and Comparison Methods section of your paper, you mentioned "all methods utilize U-Net as the backbone". Does this mean the UNet for all comparison methods contians ConvMixer layer?

哥 为什么我快训练完了在结尾会报这个错 搞不懂呜呜