Could not run with video trained models

[vxfranky]

Please refer the log as following. I've tried both in GUI and shell, given the same error. Other models without 'video' in filename could run properly.

D:\DeepMosaics_0.5.1_gpu\core\torchvision\__init__.py:26: UserWarning: You are importing torchvision within its own root folder (D:\DeepMosaics_0.5.1_gpu\core). This is not expected to work and may give errors. Please exit the torchvision project source and relaunch your python interpreter.
segment parameters: 12.4M
netG parameters: 26.65M
--------------------ERROR--------------------
--------------Environment--------------
DeepMosaics: 0.5.1
Python: 3.7.3 (default, Apr 24 2019, 15:29:51) [MSC v.1915 64 bit (AMD64)]
Pytorch: 1.7.1
OpenCV: 4.1.2
Platform: Windows-10-10.0.22621-SP0
--------------BUG--------------
Error Type: <class 'RuntimeError'>
Error(s) in loading state_dict for BVDNet:
        Missing key(s) in state_dict: "encoder3d.model.0.weight_orig", "encoder3d.model.0.weight", "encoder3d.model.0.weight_u", "encoder3d.model.0.bias", "encoder3d.model.0.weight_orig", "encoder3d.model.0.weight_u", "encoder3d.model.0.weight_v", "encoder3d.model.2.weight_orig", "encoder3d.model.2.weight", "encoder3d.model.2.weight_u", "encoder3d.model.2.bias", "encoder3d.model.2.weight_orig", "encoder3d.model.2.weight_u", "encoder3d.model.2.weight_v", "encoder3d.model.4.weight_orig", "encoder3d.model.4.weight", "encoder3d.model.4.weight_u", "encoder3d.model.4.bias", "encoder3d.model.4.weight_orig", "encoder3d.model.4.weight_u", "encoder3d.model.4.weight_v", "encoder3d.model.6.weight_orig", "encoder3d.model.6.weight", "encoder3d.model.6.weight_u", "encoder3d.model.6.bias", "encoder3d.model.6.weight_orig", "encoder3d.model.6.weight_u", "encoder3d.model.6.weight_v", "encoder2d.model.1.weight_orig", "encoder2d.model.1.weight", "encoder2d.model.1.weight_u", "encoder2d.model.1.bias", "encoder2d.model.1.weight_orig", "encoder2d.model.1.weight_u", "encoder2d.model.1.weight_v", "encoder2d.model.4.weight_orig", "encoder2d.model.4.weight", "encoder2d.model.4.weight_u", "encoder2d.model.4.bias", "encoder2d.model.4.weight_orig", "encoder2d.model.4.weight_u", "encoder2d.model.4.weight_v", "encoder2d.model.7.weight_orig", "encoder2d.model.7.weight", "encoder2d.model.7.weight_u", "encoder2d.model.7.bias", "encoder2d.model.7.weight_orig", "encoder2d.model.7.weight_u", "encoder2d.model.7.weight_v", "encoder2d.model.10.weight_orig", "encoder2d.model.10.weight", "encoder2d.model.10.weight_u", "encoder2d.model.10.bias", "encoder2d.model.10.weight_orig", "encoder2d.model.10.weight_u", "encoder2d.model.10.weight_v", "blocks.0.conv_block.1.weight_orig", "blocks.0.conv_block.1.weight", "blocks.0.conv_block.1.weight_u", "blocks.0.conv_block.1.bias", "blocks.0.conv_block.1.weight_orig", "blocks.0.conv_block.1.weight_u", "blocks.0.conv_block.1.weight_v", "blocks.0.conv_block.4.weight_orig", "blocks.0.conv_block.4.weight", "blocks.0.conv_block.4.weight_u", "blocks.0.conv_block.4.bias", "blocks.0.conv_block.4.weight_orig", "blocks.0.conv_block.4.weight_u", "blocks.0.conv_block.4.weight_v", "blocks.1.conv_block.1.weight_orig", "blocks.1.conv_block.1.weight", "blocks.1.conv_block.1.weight_u", "blocks.1.conv_block.1.bias", "blocks.1.conv_block.1.weight_orig", "blocks.1.conv_block.1.weight_u", "blocks.1.conv_block.1.weight_v", "blocks.1.conv_block.4.weight_orig", "blocks.1.conv_block.4.weight", "blocks.1.conv_block.4.weight_u", "blocks.1.conv_block.4.bias", "blocks.1.conv_block.4.weight_orig", "blocks.1.conv_block.4.weight_u", "blocks.1.conv_block.4.weight_v", "blocks.2.conv_block.1.weight_orig", "blocks.2.conv_block.1.weight", "blocks.2.conv_block.1.weight_u", "blocks.2.conv_block.1.bias", "blocks.2.conv_block.1.weight_orig", "blocks.2.conv_block.1.weight_u", "blocks.2.conv_block.1.weight_v", "blocks.2.conv_block.4.weight_orig", "blocks.2.conv_block.4.weight", "blocks.2.conv_block.4.weight_u", "blocks.2.conv_block.4.bias", "blocks.2.conv_block.4.weight_orig", "blocks.2.conv_block.4.weight_u", "blocks.2.conv_block.4.weight_v", "blocks.3.conv_block.1.weight_orig", "blocks.3.conv_block.1.weight", "blocks.3.conv_block.1.weight_u", "blocks.3.conv_block.1.bias", "blocks.3.conv_block.1.weight_orig", "blocks.3.conv_block.1.weight_u", "blocks.3.conv_block.1.weight_v", "blocks.3.conv_block.4.weight_orig", "blocks.3.conv_block.4.weight", "blocks.3.conv_block.4.weight_u", "blocks.3.conv_block.4.bias", "blocks.3.conv_block.4.weight_orig", "blocks.3.conv_block.4.weight_u", "blocks.3.conv_block.4.weight_v", "decoder.0.convup.2.weight_orig", "decoder.0.convup.2.weight", "decoder.0.convup.2.weight_u", "decoder.0.convup.2.bias", "decoder.0.convup.2.weight_orig", "decoder.0.convup.2.weight_u", "decoder.0.convup.2.weight_v", "decoder.1.convup.2.weight_orig", "decoder.1.convup.2.weight", "decoder.1.convup.2.weight_u", "decoder.1.convup.2.bias", "decoder.1.convup.2.weight_orig", "decoder.1.convup.2.weight_u", "decoder.1.convup.2.weight_v", "decoder.2.convup.2.weight_orig", "decoder.2.convup.2.weight", "decoder.2.convup.2.weight_u", "decoder.2.convup.2.bias", "decoder.2.convup.2.weight_orig", "decoder.2.convup.2.weight_u", "decoder.2.convup.2.weight_v", "decoder.4.weight", "decoder.4.bias".
        Unexpected key(s) in state_dict: "encoder_2d.model.1.weight", "encoder_2d.model.1.bias", "encoder_2d.model.5.weight", "encoder_2d.model.5.bias", "encoder_2d.model.9.weight", "encoder_2d.model.9.bias", "encoder_2d.model.13.weight", "encoder_2d.model.13.bias", "encoder_2d.model.17.weight", "encoder_2d.model.17.bias", "encoder_3d.inconv.conv.0.weight", "encoder_3d.inconv.conv.0.bias", "encoder_3d.down1.conv.0.weight", "encoder_3d.down1.conv.0.bias", "encoder_3d.down2.conv.0.weight", "encoder_3d.down2.conv.0.bias", "encoder_3d.down3.conv.0.weight", "encoder_3d.down3.conv.0.bias", "encoder_3d.down4.conv.0.weight", "encoder_3d.down4.conv.0.bias", "decoder_2d.model.0.conv_block.1.weight", "decoder_2d.model.0.conv_block.1.bias", "decoder_2d.model.0.conv_block.5.weight", "decoder_2d.model.0.conv_block.5.bias", "decoder_2d.model.1.conv_block.1.weight", "decoder_2d.model.1.conv_block.1.bias", "decoder_2d.model.1.conv_block.5.weight", "decoder_2d.model.1.conv_block.5.bias", "decoder_2d.model.2.conv_block.1.weight", "decoder_2d.model.2.conv_block.1.bias", "decoder_2d.model.2.conv_block.5.weight", "decoder_2d.model.2.conv_block.5.bias", "decoder_2d.model.3.conv_block.1.weight", "decoder_2d.model.3.conv_block.1.bias", "decoder_2d.model.3.conv_block.5.weight", "decoder_2d.model.3.conv_block.5.bias", "decoder_2d.model.4.conv_block.1.weight", "decoder_2d.model.4.conv_block.1.bias", "decoder_2d.model.4.conv_block.5.weight", "decoder_2d.model.4.conv_block.5.bias", "decoder_2d.model.5.conv_block.1.weight", "decoder_2d.model.5.conv_block.1.bias", "decoder_2d.model.5.conv_block.5.weight", "decoder_2d.model.5.conv_block.5.bias", "decoder_2d.model.6.conv_block.1.weight", "decoder_2d.model.6.conv_block.1.bias", "decoder_2d.model.6.conv_block.5.weight", "decoder_2d.model.6.conv_block.5.bias", "decoder_2d.model.7.conv_block.1.weight", "decoder_2d.model.7.conv_block.1.bias", "decoder_2d.model.7.conv_block.5.weight", "decoder_2d.model.7.conv_block.5.bias", "decoder_2d.model.8.conv_block.1.weight", "decoder_2d.model.8.conv_block.1.bias", "decoder_2d.model.8.conv_block.5.weight", "decoder_2d.model.8.conv_block.5.bias", "decoder_2d.model.9.weight", "decoder_2d.model.9.bias", "decoder_2d.model.12.weight", "decoder_2d.model.12.bias", "decoder_2d.model.15.weight", "decoder_2d.model.15.bias", "decoder_2d.model.18.weight", "decoder_2d.model.18.bias", "decoder_2d.model.22.weight", "decoder_2d.model.22.bias", "merge1.conv.1.weight", "merge1.conv.1.bias".
<FrameSummary file deepmosaic.py, line 77 in <module>>
<FrameSummary file deepmosaic.py, line 41 in main>
<FrameSummary file models\loadmodel.py, line 56 in video>
<FrameSummary file torch\nn\modules\module.py, line 1052 in load_state_dict>
Please press any key to exit.

[gibrahim-cr]

I had the same issue exactly! The only video cleaning pth that works on video was clean_youknow_resnet_9blocks.pth but it is pretty crappy and really meant for images. IMHO based on my limited knowledge of python and no knowledge of AI and of course based on the errors and what works/doesn't I'm pretty sure it compatibility issues with torch. I even tried using conda to set up a vert env to control package versions, same issue. The only thing I haven't tried is to spin up a vanilla VM and try it there which I WILL EVENTUALLY DO!