gaetandi / chexpert Goto Github PK

Hello @gaetandi!
First of all, I want to thank you for uploading this awesome repo. I just started to learn more about ML and I think I can learn a lot from your repo.
Unfortunately, I get the error "This does not look like a tar archive" directly in my terminal when unpacking the file "model_ones_3epoch_densenet.tar".
It may well be that I make a simple beginner mistake, but unfortunately, I can not find a solution. I would be very happy about possible solutions.

OS: Ubuntu 18.04.2 LTS

Do you do anything to manage class imbalance? Weighted random sampler or similar?

From where to get train and validation CSV Datasets?
Provide link

Hi, thank you for posting this code. I am trying to do fine tuning on similar data using your model and weights but am running into some issues when I try to edit the final layer to fit 2 classes. If I try ```
model.classifier = nn.Sequential(
nn.Linear(1024, num_classes),
nn.Sigmoid()

I end up with just a 2nd classifier layer when I print(model). 

It seems like children removal doesn't work for densenets.. 
`model = nn.Sequential(*list(model.children())[:-1])`

How can I replace the final layer? Thank you

as titile

Hi, can you please tell me how to generate all the 14 labels' heatmap for the given image? Also, it would be great if I could get some explanation of how the heatmap is calculated.