xhp-hust-2018-2011 / s-dcnet Goto Github PK

I could not download data and model. Can you help me because there are some problems.

Could you give the codes of generating density map and the code snippet of get CMax?

Hi @xhp-hust-2018-2011 @poppinace

Please allow me to ask, in your paper section 3.2 and in your code, W1 has the same size of C1 with a shape of (2,2), where in your Section 5.2(Figure 8), the input crop seems have the same size compared with W1 visualization.

Is this W1 in section 5.2 the same as W1 in section 3.2? If yes, why could they have the same size? How you visualize W1 in this case?

Thank you for your work! If I want to infer single Image ,how can i do it?

Thanks first for the beautiful work you did.

When aplying the code on large amount of images, the memory use increases after efvery processing of one image until crash of system.
I tried to delete some variable after use without any effect
how to solve this memory issue?

For the line "cl2h[:,:,rx::rate,ry::rate] = clow*norm" in merge_func.py, I am getting the following runtime error "Expanded size of Tensor must match existing dimension for non-singleton dimension". This happens when the size of chigh is not exactly twice the size of clow. Specifically this happens when clow dimensions are odd numbers. Can you provide a solution on how to fix this issue?

How to process the frames in order?

cd：root/S-DCNet-master/Network/merge_func.PY/ line23
for rx in range(rate):
for ry in range(rate):
cl2h[:,:,rx::rate,ry::rate] = clow*norm
Report an error at this location，
【RuntimeError: The expanded size of the tensor (9) must match the existing size (8) at non-singleton dimension 3. Target sizes: [1, 1, 6, 9]. Tensor sizes: [5, 8]】
Can you tell me why？

Hello, is it possible to apply for our original dataset?
If so, please tell me how to generate density map or other necessary preparation.
Thanks in advnace.

Hi @xhp-hust-2018-2011

Thank you for your excellent work.

I want to ask questions for the shanghai tech dataset you used. In your implementation("Loaddata.py"), ground true files for training mat has a field called "crop_gtdens" and for test mat you have a field called "density_map". Both of them are not available in original dataset.

Can you explain what "crop_gtdens" is? Maybe density map for cropped image?

Thank you

hi,thanks for sharing the code,but i have a few question,
1)what is the input of cross-entropy loss?one is the CLS0~CLS1 in div_res and the other one is the label_indice?but ,the CLS0,CLS1,CLS2 have the different shape ,how can the be computed for loss?
2)what is the input of L1 loss?when division time is 2,one is the DIV2 in merge_res,and what is the other one?
3)in data augmentation,you said crop 4 corners and randomly crop 5 patch,did you concat these 9 patch into a 93h*w tensor and send into the network?
looking forward of your reply,thanks a lot!!

In algorithm 1, the paper mentioned "Integrate over DIV_N to obtain the image count C". How to integrate over DIV_N(an array)? Any reference?

Thank you for your interesting work.
I would like to know if you have trained the model on UCF-QNRF?

How do you generate the density map? You didn't share your code about that. Can you explain to help?