Provided model fails to load about monodepth-pytorch

[NikolasEnt]

Yes, you are right, we are currently updating the model after fixing several bugs in architecture. That is why, the provided model weights should not fit the model.
We will update the provided model as soon as possible. I apologize for the inconvenience.

[stashvala]

No problem, just wanted to try out the model. I also had to switch torch.Resize to torch.Scale in transforms.py. Thanks for the quick response, good luck with fixing!

[xiaohedu]

Yes, you are right, we are currently updating the model after fixing several bugs in architecture. That is why, the provided model weights should not fit the model.
We will update the provided model as soon as possible. I apologize for the inconvenience.

look forward to updating as soon as possible

[voeykovroman]

@xiaohedu We will provide pretrained model in a couple days. But until that you can try to train this model yourself if you have KITTI dataset.

[xiaohedu]

@xiaohedu We will provide pretrained model in a couple days. But until that you can try to train this model yourself if you have KITTI dataset.
thank you.

[voeykovroman]

While we are training a new model (have technical issues, probably takes one more week) you can switch to commit 5664039 and use our old model provided in README (demo video in README was made using exactly this model).
But for correct work you should replace ImageLoader with KittiLoader in main_monodepth_pytorch.py in line 160. And also make relevant changes in paths to images and to model in demo notebook.

[NikolasEnt]

@stashvala @xiaohedu Ok, we updated the code in the master branch, fixed several bugs and provided a new pretrained model along with the updated jupyter notebook for the training recreation. Hope, it helps.