The network creates 3 outputs. Is there any additional computation involved between the loss function and the 3 net outputs? Thanks

hi~

When I run this program with "bash CHR/runme.sh"，the output in terminal is :

Traceback (most recent call last):
File "/usr/lib/python3.6/runpy.py", line 193, in _run_module_as_main
"main", mod_spec)
File "/usr/lib/python3.6/runpy.py", line 85, in _run_code
exec(code, run_globals)
File "/home/yg/Desktop/workspace/CHR/CHR/CHR/CHR/main.py", line 110, in
main_ray()
File "/home/yg/Desktop/workspace/CHR/CHR/CHR/CHR/main.py", line 83, in main_ray
train_dataset = XrayClassification(args.data, 'train')
File "/home/yg/Desktop/workspace/CHR/CHR/CHR/CHR/ray.py", line 95, in init
os.makedirs(path_csv)
File "/usr/lib/python3.6/os.py", line 210, in makedirs
makedirs(head, mode, exist_ok)
File "/usr/lib/python3.6/os.py", line 210, in makedirs
makedirs(head, mode, exist_ok)
File "/usr/lib/python3.6/os.py", line 210, in makedirs
makedirs(head, mode, exist_ok)
[Previous line repeated 2 more times]
File "/usr/lib/python3.6/os.py", line 220, in makedirs
mkdir(name, mode)
PermissionError: [Errno 13] Permission denied: '/DATA'

In fact , I have not folder named '/DATA'.
What is the purpose of this folder？ And how can I use dataset(SIXray) in this project?

thx!

Hi @MeioJane ,

Can you please upload the PyTorch trained model weights?
This will be helpful for reproducibility purposes.

Thanks. :)

To verify that CHR is not over-tuned to image classification, you attach the class activation map (CAM) .Where is this part of the code

Could you tell me where is the *.yml file ?
I can not find it~
Thanks very much !!

I want to test own dataset using the CHR weight file.
But I can't download it on BaiduDrive.

So, Could you upload it on google-drive or onedrive?

Hi, I saw your research result. Looks good, thank you.

1. Can I request the pre-trained model?
2. And I couldn't join the Baidu, so could you share the download other link?

Thank you

Hello, I want to use MulticlassTop5Engine(Class) to measure accuracy.
But get some errors:
"input_var = torch.autograd.Variable(self.state['input'])
TypeError: Variable data has to be a tensor, but got list"
But when using MultiLabelMAPEngine(Class), not those errors.

Can you add the code of how to run class MluticlassTop5Engine?
Thanks!

If someone downloaded the dataset, could explain the way or share it (not on Baidu), please?

It is mentioned that CHR approach used only one sample to update the learnable parameters, similar to stochastic gradient descent. I am wondering if, during the experiments, you also trained the plain CNN models in the same way. Did you also use only one sample for training plain models (say ResNet50, DenseNet, etc.)? If not, what is the batch size? More generally, can you state the training set up for both the plain CNN and CHR method? (Learning rate, optimizer, batch size, number of epochs, etc.) It is not clear whether the default values in your code were the exact values used in training. Also, did you use pretrained weights for the backbones?

thanks for the great work. I see that the loss of the netwok is a self implemented binary cross entropy loss parameterized by class weights. (I guess the weights are used to adress the imbalanced data labels?). I read the code in main.py, but I didn't see the weights take part in any computation
-(target*torch.log(input+eps)+(1-target)*torch.log(1-input+eps))
Would this be updated later? Thanks

Actually it seems not defined globally within the hole projects, so maybe it's a piece of testing code before release?

Thanks for your work! but i have a problem,i dont know how to do solve it.
I have downloaded Resnet101-CHR model. I want to test image,how to do this work?

Hi, it's a great job that you made such a dataset. But I meet a problem that I could onlt downlad 1.4k annotations, could you please upload the full annotation files? Thanks.

can not found any code for parse xml.

How to found it,

Are the results on your paper based on the mAP calculated with difficult examples set as True? It seems that in your code, this was set to true.

Hello, @MeioJane,

First of all let me congratulate you. As a Computer Vision enthusiast, I admire the work you've made in this research. I would very much like to reproduce your results and build upon your amazing work. But, in order to do that, I would need your trained model weights, as I can't afford to retrain the model from scratch right now. I tried to download the weights using the link you've provided (from Baidu) but I was unsuccessful. Could you please share the file with me, via another link?

Thanks in advance. My institutional e-mail is [email protected].

thanks for your work, could you upload the pretrained model? thanks.

When I run the program,Instead of using the GPU, the memory utilization keeps increasing until it is 100%.How to solve this problem?Thank you.