gjy3035 / gcc-sfcn Goto Github PK

from .resSFCN import resSFCN as net
SystemError: Parent module '' not loaded, cannot perform relative import

Hi ,
I am using test.py with (torch 1.0 python 3.5.4 ),after i changed the model path to downloaded "Pretrained_GCC.pth" .It seems this pth's state_dict mismatchs CrowdCounter. Where Can I find the right pretrain pth for CrowdCounter? or Did i miss things?Please let me know.

Hello, I was training because the computer has only one gpu0 reason to fail, and I have been on a tight schedule recently, and I want to run the test directly. Can you email me the model(all_ep_178_mae_74.1_mse_131.0.pth) you need for the test?The model is the line 36 in test.py.Thanks!

Hi，thanks for your nice work. Could you show me where is the SSIM loss in the code？

Hello, have you counted the number of parameters and flops of this model?I want to count these two parameters to see if I can use this model on the embedded platform, but I have some problems. @gjy3035

Hi, many thanks for your work.

I noticed that when doing forward prediction in test.py, result from network will be divided by 100.0:
pred = np.sum(pred_map)/100.0

This confuses me because I don't see any division like this in loss definition in train.py or CC.py. Do you use different processed groundtruth data for training and validation?

Hi, Junyu, have you tried to pre-train the supervised SFCN model using translated GCC datasets via SE Cycle GAN and then, fine-tune the model with real images? This may work well, i think.

When I want to train with my own data set, I get the following error
D:\Program Files\Anaconda3\envs\lei\envs\py36-pytorch\python.exe" D:/zgy/GCC/GCC-SFCN-master/train.py
Traceback (most recent call last):
File “D:/zgy/GCC/GCC-SFCN-master/train.py”, line 250, in
main()
File “D:/zgy/GCC/GCC-SFCN-master/train.py”, line 100, in main
i_tb = train(train_loader, net, optimizer, epoch, i_tb)
File “D:/zgy/GCC/GCC-SFCN-master/train.py”, line 120, in train
for i, data in enumerate(train_loader, 0):
File “D:\Program Files\Anaconda3\envs\lei\envs\py36-pytorch\lib\site-packages\torch\utils\data\dataloader.py”, line 286, in next
return self._process_next_batch(batch)
File “D:\Program Files\Anaconda3\envs\lei\envs\py36-pytorch\lib\site-packages\torch\utils\data\dataloader.py”, line 307, in _process_next_batch
raise batch.exc_type(batch.exc_msg)
AssertionError: Traceback (most recent call last):
File “D:\Program Files\Anaconda3\envs\lei\envs\py36-pytorch\lib\site-packages\torch\utils\data\dataloader.py”, line 57, in _worker_loop
samples = collate_fn([dataset[i] for i in batch_indices])
File “D:\Program Files\Anaconda3\envs\lei\envs\py36-pytorch\lib\site-packages\torch\utils\data\dataloader.py”, line 57, in
samples = collate_fn([dataset[i] for i in batch_indices])
File “D:\zgy\GCC\GCC-SFCN-master\datasets\UCF_QNRF.py”, line 40, in getitem
img, den = self.main_transform(img,den)
File “D:\zgy\GCC\GCC-SFCN-master\misc\transforms.py”, line 16, in call
img, mask = t(img, mask)
File “D:\zgy\GCC\GCC-SFCN-master\misc\transforms.py”, line 50, in call
assert img.size == mask.size
AssertionError

Since I don't get your point why we have to generate the density maps or resize the images with the segmentation. I would appreciate it if you could get this stuff cleared or adjust your code. Thx!

Is it a real-time network?If not， what is the speed? thx.

result in the paper of shanghaitech partA is 64.8/107.5 can you give me the code of this part or some setting of your experiment .i have run the code of C-3 resnet50 but i got partA 77.3/124.5 with batchsize=6
crop image size of batch min size

error

Hello, can I simply enter a picture and get a prediction result? @gjy3035

Can this program can run in python3.6 and pytorch1.0.1？And I want to use VGG16 as the skeleton.Is that feasible?

Excellent work!
Density map is obtained, but how to further generate corresponding bounding boxes on image?
Any idea? thx

in test.py file:
denname = './den/' + filename_no_ext + '.csv'
den = pd.read_csv(denname, sep=',',header=None)

what does the .csv file used for ?

Hello Junyu
I try to translate images with the CycleGAN code you provide with the filtered synthetic images for each real dataset. But the results I get (mae or mse) is even worse than the results given by the model trained with the original images. Could you provide the transferred images. And it would be great if you can provide all the code, parameters and input images to train CycleGAN.

Is pdf of the paper available? I couldn't find it.

The len of pretrained model's state_dict is 488 and that of resSFCN is 582.
If I want to finetune the model, what the details should I notice? For example, in shanghaiB, how to preprocess images and how to set optimizer?

What is the configuration of your computer for experiment?
Can the code run in GPU[0] only?

Hii! Does the code have the cycle-gan part of the paper? Thanks!

now GPU P100 test，200ms/per，but it is slower for me，how to do ？？

Runtimeerror:Error(s) in loading state_dict for ResNet:
Missing key(s) in state_dict:
Unexpected key(s) in state_dict:
I have encountered the above problem.Weight and residual network do not correspond.

I would like to do a quick experiment with vgg network, I would appreciate if you could help on it.

The GCC dataset is great! but is the photorealistic GCC dataset available? The one transformed by CycleGAN or SSIM?

Thanks.

Can I know how much time will it take to train the SOTA？ Thanks！

After getting the density map, how can I get the counting number results?

Would it be possible for you to release the final model trained on the shanghaitech B dataset? I wanted it to evaluate some of my own images on a custom dataset. Thanks!

Hi,Thanks for your sharing.I met a problem as follows when i run train.py
train.py:125: UserWarning: volatile was removed and now has no effect. Use with torch.no_grad():instead. img = Variable(img, volatile=True).cuda() train.py:126: UserWarning: volatile was removed and now has no effect. Usewith torch.no_grad(): instead. gt_map = Variable(gt_map, volatile=True).cuda() Traceback (most recent call last): File "train.py", line 205, in <module> main() File "train.py", line 81, in main validate(val_loader, net, epoch, restore_transform) File "train.py", line 141, in validate gt_count_tmp = gt_count[i_img] IndexError: index 1 is out of bounds for axis 0 with size 1

This happend in validate(),is there something wrong in this function? I have print the gt_count.shape,it equals to (1,),and pred_map.shape equals to (768,1024),and i don't understand why use gt_count_tmp=gt_count[0] in line 141 of train.py.
Thanks for your reply.

Hi, I was running test.py on single GPU but got this error. Do you know what is wrong? Thank you so much.

Ubuntu 16.04
python 2.7
pytorch 0.4.0
CUDA 9.0
CUDNN 7.0

Traceback (most recent call last):
  File "test.py", line 185, in <module>
    main()
  File "test.py", line 54, in main
    test(file_list[0], model_path)
  File "test.py", line 111, in test
    pred_map = net.test_forward(img)
  File "/home/work/zhanghaoxu/GCC-SFCN/models/CC.py", line 43, in test_forward
    density_map = self.CCN(img)            
  File "/software/conda/envs/zhanghaoxu/lib/python2.7/site-packages/torch/nn/modules/module.py", line 491, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/work/zhanghaoxu/GCC-SFCN/models/resSFCN.py", line 44, in forward
    x = self.frontend(x)
  File "/software/conda/envs/zhanghaoxu/lib/python2.7/site-packages/torch/nn/modules/module.py", line 491, in __call__
    result = self.forward(*input, **kwargs)
  File "/software/conda/envs/zhanghaoxu/lib/python2.7/site-packages/torch/nn/modules/container.py", line 91, in forward
    input = module(input)
  File "/software/conda/envs/zhanghaoxu/lib/python2.7/site-packages/torch/nn/modules/module.py", line 491, in __call__
    result = self.forward(*input, **kwargs)
  File "/software/conda/envs/zhanghaoxu/lib/python2.7/site-packages/torch/nn/modules/container.py", line 91, in forward
    input = module(input)
  File "/software/conda/envs/zhanghaoxu/lib/python2.7/site-packages/torch/nn/modules/module.py", line 491, in __call__
    result = self.forward(*input, **kwargs)
  File "/software/conda/envs/zhanghaoxu/lib/python2.7/site-packages/torchvision/models/resnet.py", line 83, in forward
    out = self.conv2(out)
  File "/software/conda/envs/zhanghaoxu/lib/python2.7/site-packages/torch/nn/modules/module.py", line 491, in __call__
    result = self.forward(*input, **kwargs)
  File "/software/conda/envs/zhanghaoxu/lib/python2.7/site-packages/torch/nn/modules/conv.py", line 301, in forward
    self.padding, self.dilation, self.groups)
RuntimeError: CUDNN_STATUS_INTERNAL_ERROR

Hello, I tried to download the GCC dataset from your project website but I failed.
How can I get the GCC dataset ?
And, I want to use it as the pretrain dataset for my model, how to deal with the data?
Thanks a lot ~
Best wishes.