zijundeng / bdrar Goto Github PK

Hi @zijundeng ,
I have loaded your pre-trained model and inference, but I got Ber = 5.20. I have ensured that my evaluation code is right, while I used your network, data loader and other.
I don't know whether is there some problems with your pre-trained model or other.
Thanks.

I cannot load the pre-trained model when run infer.py. There are many mismatched parameters.
For example, in the refine3_h2l module of the BDRAR network, the parameters'name in network is
################
refine3_h2l.0.weight
refine3_h2l.1.weight
refine3_h2l.1.bias
refine3_h2l.3.weight
refine3_h2l.4.weight
refine3_h2l.4.bias
refine3_h2l.6.weight
refine3_h2l.7.weight
refine3_h2l.7.bias
refine3_l2h.0.weight
refine3_l2h.1.weight
refine3_l2h.1.bias
refine3_l2h.3.weight
refine3_l2h.4.weight
refine3_l2h.4.bias
refine3_l2h.6.weight
refine3_l2h.7.weight
refine3_l2h.7.bias
################

However, in the pre-trained models, the corresponding parameters'name is:
################
refine3_hl.0.weight
refine3_hl.1.weight
refine3_hl.1.bias
refine3_hl.1.running_mean
refine3_hl.1.running_var
refine3_hl.3.weight
refine3_hl.4.weight
refine3_hl.4.bias
refine3_hl.4.running_mean
refine3_hl.4.running_var
refine3_hl.6.weight
refine3_hl.7.weight
refine3_hl.7.bias
refine3_hl.7.running_mean
refine3_hl.7.running_var
refine3_lh.0.weight
refine3_lh.1.weight
refine3_lh.1.bias
refine3_lh.1.running_mean
refine3_lh.1.running_var
refine3_lh.3.weight
refine3_lh.4.weight
refine3_lh.4.bias
refine3_lh.4.running_mean
refine3_lh.4.running_var
refine3_lh.6.weight
refine3_lh.7.weight
refine3_lh.7.bias
refine3_lh.7.running_mean
refine3_lh.7.running_var
################

Thanks for sharing the whole training and testing codes. So could you further share the pre-trained BDRAR model for salient object detection?

I used pretrained model uploaded in the link and run inference for SBU test set.
These results do not match with the published results.
Why is such a case?

Hi ,
I used pytorch-0.4.0 , python2 will make a error:
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation.
Use pytorch-0.3 to run normally.

Hello, first thanks for sharing the code and the trained network.
I downloaded 3000.pth that you provided, and wanted to do further fine tuning on other shadow dataset.
When I run train.py I saw that the file 3000_optim.pth is missing,
Can you upload it as well?

Many thanks,
Tamir

Hello there,

I am trying to fine tune the model you shared with a new dataset, but I can't get it running due to the following error:

lib/python2.7/site-packages/torch/nn/functional.py:1749: UserWarning: Default upsampling behavior when mode=bilinear is changed to align_corners=False since 0.4.0. Please specify align_corners=True if the old behavior is desired. See the documentation of nn.Upsample for details. "See the documentation of nn.Upsample for details.".format(mode)) Traceback (most recent call last): File "train.py", line 168, in <module> main() File "train.py", line 95, in main train(net, optimizer) File "train.py", line 136, in train loss.backward() File "/home/eloyroura/anaconda3/envs/BDRAR/lib/python2.7/site-packages/torch/tensor.py", line 93, in backward torch.autograd.backward(self, gradient, retain_graph, create_graph) File "/home/eloyroura/anaconda3/envs/BDRAR/lib/python2.7/site-packages/torch/autograd/__init__.py", line 89, in backward allow_unreachable=True) # allow_unreachable flag RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation

Any hint of what i am doing wrong?

At the beginning, I encountered a problem like this:
"RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation"
So, I solved it by changing

for m in self.modules():
    if isinstance(m, nn.ReLU) or isinstance(m, nn.Dropout):
        m.inplace = True

to be

for m in self.modules():
    if isinstance(m, nn.ReLU):
        m.inplace = True

But a new problem occurs , "RuntimeError: cuda runtime error (2) : out of memory at /pytorch/aten/src/THC/generic/THCStorage.cu:58".
Can't a single 1080 ti support this model?

I try to run train.py,then error happen.Can anyone help?

Traceback (most recent call last):
File "D:\Program Files (x64)\PyCharm 2021.1.3\plugins\python\helpers\pydev\pydevd.py", line 1483, in exec
pydev_imports.execfile(file, globals, locals) # execute the script
File "D:\Program Files (x64)\PyCharm 2021.1.3\plugins\python\helpers\pydev_pydev_imps_pydev_execfile.py", line 18, in execfile
exec(compile(contents+"\n", file, 'exec'), glob, loc)
File "F:/20220410_BDRAR-master/train1.py", line 149, in
main()
File "F:/20220410_BDRAR-master/train1.py", line 78, in main
train(net, optimizer)
File "F:/20220410_BDRAR-master/train1.py", line 117, in train
loss.backward()
File "D:\anaconda3\envs\BDRAR\lib\site-packages\torch\tensor.py", line 245, in backward
torch.autograd.backward(self, gradient, retain_graph, create_graph, inputs=inputs)
File "D:\anaconda3\envs\BDRAR\lib\site-packages\torch\autograd_init.py", line 147, in backward
allow_unreachable=True, accumulate_grad=True) # allow_unreachable flag
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [1, 8, 128, 128]], which is output 0 of ReluBackward1, is at version 2; expected version 1 instead. Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True).
python-BaseException

Hi,
My GPU memory is enough, 11GB, but I run the training code and obtain the following errors:

/usr/local/lib/python2.7/dist-packages/torch/nn/functional.py:1890: UserWarning: nn.functional.upsample is deprecated. Use nn.functional.interpolate instead.
warnings.warn("nn.functional.upsample is deprecated. Use nn.functional.interpolate instead.")
/usr/local/lib/python2.7/dist-packages/torch/nn/functional.py:1961: UserWarning: Default upsampling behavior when mode=bilinear is changed to align_corners=False since 0.4.0. Please specify align_corners=True if the old behavior is desired. See the documentation of nn.Upsample for details.
"See the documentation of nn.Upsample for details.".format(mode))
/usr/local/lib/python2.7/dist-packages/torch/nn/functional.py:1006: UserWarning: nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.
warnings.warn("nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.")
Traceback (most recent call last):
File "train.py", line 152, in
main()
File "train.py", line 81, in main
train(net, optimizer)
File "train.py", line 112, in train
loss4_h2l = bce_logit(predict4_h2l, labels)
File "/usr/local/lib/python2.7/dist-packages/torch/nn/modules/module.py", line 477, in call
result = self.forward(*input, **kwargs)
File "/usr/local/lib/python2.7/dist-packages/torch/nn/modules/loss.py", line 573, in forward
reduction=self.reduction)
File "/usr/local/lib/python2.7/dist-packages/torch/nn/functional.py", line 1651, in binary_cross_entropy_with_logits
loss = input - input * target + max_val + ((-max_val).exp() + (-input - max_val).exp()).log()
RuntimeError: CUDA error: out of memory

I want to know how to solve is problem. As the same time, I try to change the batch size 8 to a smaller batch size 4, but it does not work, and outputs the following errors:

/usr/local/lib/python2.7/dist-packages/torch/nn/functional.py:1890: UserWarning: nn.functional.upsample is deprecated. Use nn.functional.interpolate instead.
warnings.warn("nn.functional.upsample is deprecated. Use nn.functional.interpolate instead.")
/usr/local/lib/python2.7/dist-packages/torch/nn/functional.py:1961: UserWarning: Default upsampling behavior when mode=bilinear is changed to align_corners=False since 0.4.0. Please specify align_corners=True if the old behavior is desired. See the documentation of nn.Upsample for details.
"See the documentation of nn.Upsample for details.".format(mode))
/usr/local/lib/python2.7/dist-packages/torch/nn/functional.py:1006: UserWarning: nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.
warnings.warn("nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.")
Traceback (most recent call last):
File "train.py", line 153, in
main()
File "train.py", line 82, in main
train(net, optimizer)
File "train.py", line 121, in train
loss.backward()
File "/usr/local/lib/python2.7/dist-packages/torch/tensor.py", line 93, in backward
torch.autograd.backward(self, gradient, retain_graph, create_graph)
File "/usr/local/lib/python2.7/dist-packages/torch/autograd/init.py", line 90, in backward
allow_unreachable=True) # allow_unreachable flag
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation

So can you help me solve this issue? Very thank you.

when i run this code,it show Attempting to deserialize object on CUDA device 1 but torch.cuda.device_count() is 1. Please use torch.load with map_location to map your storages to an existing device.Do you know how to solve this ?

code crashes after BDRAR().cuda()

Hi guys,

Thanks so much for open sourcing your code. However without any license no-one can run/develop from your code. Could you please add a license too, with the code?

If you want it to be fully open-source, I would suggest an Apache 2.0 License: https://tldrlegal.com/license/apache-license-2.0-(apache-2.0)

UCF dataset cannot be accessed, can you provide it?
Thank You！！

thank you for your great job!
but i want to remove the shadow,how to do this?
Could you give me some ideas?