woodman718 / capsnets Goto Github PK

Thanks for good research and i'm interested in skin cancer lession classification task. i want to ask some questions. How to get the data set for this experiment？What do the different compressed packages of this website "https://aistudio.baidu.com/aistudio/datasetdetail/151696/0" represent?

Can you publish the code you use to perform data augmentation to handle imbalanced and split datasets?

At this point this notebook just doesn't continue running for me, after the display "[Train Epoch:[1]HAM10000 ==> Training] ..." nothing more comes, there is also no activity on GPU or CPU anymore.

I have a few questions to understand how these very good metrics came about. Do I see correctly that first the augmentations were done and then the split into the three subsets for training, validation and testing? Doesn't that lead to overfitting or mere recognition of images when augmented variants of images are split between training and testing datasets? Or was this prevented and I missed it? I think the usual approach would be to use a dataloader with Weighted Random Sampler and augmentations for training only, but leave the test dataset unchanged. How does the quality of your model (accuracy, average F1, MCC) actually look on the official test dataset? https://dataverse.harvard.edu/file.xhtml?persistentId=doi:10.7910/DVN/DBW86T/OSKJF2&version=4.0

I am really interested in your work, but I have a question that I couldn't understand in the code.
How does the resulting accuracy change when you change the test batch size ?
I can't find any parameter or operation done in testing that depends on batch size.

Im really interested in your work and really liked the result you got
Actually i have a problem
When i make the batch size is 1 on the validation dataset the accuracy drops to 33% accuracy
Do you have any recommendations to solve this issue

The number of parameters reported for CapsNet seems to be coming from torchsummary.

However, torchsummary parameter count is inaccurate for custom modules, in particular when parameters name is not "weight" or "bias" (see sksq96/pytorch-summary#65). The count does not consider network.digits.W parameters that represent ~25% of CapsNet parameters.