error occurred

[ForrestPi]

when run
python train_gaussianhead.py --config config/train_gaussianhead_N031.yaml
error blow

raceback (most recent call last):
  File "train_gaussianhead.py", line 89, in <module>
    trainer.train(0, 1000)
  File "/data0/XXX/avatar/Gaussian-Head-Avatar/lib/trainer/GaussianHeadTrainer.py", line 57, in train
    supres_images = self.supres(cropped_render_images)
  File "/home/XXX/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl
    return forward_call(*input, **kwargs)
  File "/data0/XXX/avatar/Gaussian-Head-Avatar/lib/module/SuperResolutionModule.py", line 14, in forward
    output = self.upsampler(input)
  File "/home/XXX/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl
    return forward_call(*input, **kwargs)
  File "/data0/XXX/avatar/Gaussian-Head-Avatar/lib/network/Upsampler.py", line 81, in forward
    x = self.inc(x)
  File "/home/XXX/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl
    return forward_call(*input, **kwargs)
  File "/data0/XXX/avatar/Gaussian-Head-Avatar/lib/network/Upsampler.py", line 23, in forward
    return self.double_conv(x)
  File "/home/XXX/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl
    return forward_call(*input, **kwargs)
  File "/home/XXX/anaconda3/lib/python3.8/site-packages/torch/nn/modules/container.py", line 141, in forward
    input = module(input)
  File "/home/XXX/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl
    return forward_call(*input, **kwargs)
  File "/home/XXX/anaconda3/lib/python3.8/site-packages/torch/nn/modules/conv.py", line 447, in forward
    return self._conv_forward(input, self.weight, self.bias)
  File "/home/XXX/anaconda3/lib/python3.8/site-packages/torch/nn/modules/conv.py", line 443, in _conv_forward
    return F.conv2d(input, weight, bias, self.stride,
RuntimeError: Given groups=1, weight of size [128, 32, 3, 3], expected input[1, 3, 512, 512] to have 32 channels, but got 3 channels instead

[YuelangX]

Please confirm that the output dimension of the color MLPs are equal to the input dimension of the super resolution network. You can modify it in the config file.

[ForrestPi]

@YuelangX
when run
python train_gaussianhead.py --config config/train_gaussianhead_N031.yaml
error above

[INERTIA2001]

Hello, I have the same problem with the code. Below is the configurations used, however the error stays the same. note that all the color MLPs is set to output 32 channel tensor, however after rendering, it outputs a 3 channel tensor.

Also how do I view it using SIBR viewers? since the output is a RGB image. Thank you in advanced!

[SQYmu]

same problem

[SQYmu]

the color_precomp is a 32 channel tensor, but render_images_b is a 3 channel tensor

[INERTIA2001]

Yes I also notice that, due to Gaussian rasterization, the output have 3 channels.

But after trying out different configuration, we can make it work by reducing the color MLP output to have 3 dims and the supres module to have 3 dims input.

However this deviates from the original paper where they mentioned about rendering a 32 channel image.

So I kind of dont know what to do here

[YuelangX]

@INERTIA2001 @SQYmu Do you modify "NUM_CHANNELS 3" to "NUM_CHANNELS 32" in "diff-gaussian-rasterization/cuda_rasterizer/config.h"?

[SQYmu]

@YuelangX It works, and diff_gaussian_rasterization need to reinstall because I have installed it before

[INERTIA2001]

it also works for me thanks a lot!

[INERTIA2001]

any luck visualizing gaussian using SIBR? @SQYmu

[wuxianjia1996]

@YuelangX when run python train_gaussianhead.py --config config/train_gaussianhead_N031.yaml error above

I got the same error. How did you solve it ?

[zhaowangke]

It is necessary to delete Gaussian_splatting, then pull it again, and modify /submodules/diff-Gaussian-reconstruction/cuda_rasterizer/config. h NUM-CHANNELS 32. It only works with pip uninstall diff-gaussian-rasterization then install again, It cannot be solved.

[huahangc]

And if it do not work, please delete the submodules/diff_gaussian_rasterization/build/ and submodules/diff_gaussian_rasterization/diff_gaussian_rasterization.egg-info directories and modify the NUM_CHANNEL 3 to NUM_CHANNEL 32 then use pip to install it once again.