commissarma / csrnet-pytorch Goto Github PK

Hi, I'm using CSRNet for crowd counting on world Expo'10 dataset. However, I cannot reproduce the results of the paper(8.6 MAE), I only got a MAE around 10.2. Could you please provide the implementation of world Expo'10 dataset(e.g. density map generation or how to utilize the ROI and evalutation), or at least provide some implementation details( learning rate, batch size ect.). Thank you so much!!

请问这个数据集的处理方法是什么呀

hello thinks your code ！
I'm running it .But in data_preparation/k_nearest_gaussian_kernel.py, I don't find gt .
I want to know how to set this parameter.

warnings.warn(warning.format(ret))
0%| | 0/300 [00:00<?, ?it/s]
Traceback (most recent call last):
File "G:/STUDY CODE/CSRNet/CSRNet-pytorch-master/train.py", line 29, in
for i, data in enumerate(tqdm(train_dataloader)):
File "D:\Anaconda3\envs\pytorch\lib\site-packages\tqdm\std.py", line 1129, in iter
for obj in iterable:
File "D:\Anaconda3\envs\pytorch\lib\site-packages\torch\utils\data\dataloader.py", line 615, in next
batch = self.collate_fn([self.dataset[i] for i in indices])
File "D:\Anaconda3\envs\pytorch\lib\site-packages\torch\utils\data\dataloader.py", line 615, in
batch = self.collate_fn([self.dataset[i] for i in indices])
File "G:\STUDY CODE\CSRNet\CSRNet-pytorch-master\dataset.py", line 50, in getitem
dmap = self.dmap_transform(dmap)
File "D:\Anaconda3\envs\pytorch\lib\site-packages\torchvision\transforms\transforms.py", line 61, in call
return F.to_tensor(pic)
File "D:\Anaconda3\envs\pytorch\lib\site-packages\torchvision\transforms\functional.py", line 71, in to_tensor
img = img.view(pic.size[1], pic.size[0], nchannel)
RuntimeError: shape '[1024, 832, 1]' is invalid for input of size 3407872

When I try to run train.py,I met the Error above,I can't found the reason,could you please help me.

can't find it. can you update it? thanks a lot

hi, can you show the config when you were training on the SHB dataset？ like lr and optim, etc
Or it is just the same with the config in the config.py

The test.py cannot work, do someone know how to modify it? Or update a new version?

Hi, thanks for your implementation.
It seems that the link to download the dataset from DropBox is not working.