aim-uofa / AdelaiDepth

This repo contains the projects: 'Virtual Normal', 'DiverseDepth', and '3D Scene Shape'. They aim to solve the monocular depth estimation, 3D scene reconstruction from single image problems.
Creative Commons Zero v1.0 Universal
1.06k stars 144 forks source link

Error in loading pretrained model to initialize training script #45

Closed mikacuy closed 2 years ago

mikacuy commented 2 years ago

Hi,

I am trying to initialize the training script with the provided pretrained model. In the training script, the depth_model comes with auxi_modules (module.depth_model.auxi_modules) that are not in the pretrained model files.

I was wondering how I can initialize auxi_modules so that I can use the auxi losses in my fine tuning.

Also, where in the training code is "focal_keys" and "shift_keys" trained? These were also in the pretrained model. But the number of parameters in those do not equal the missing auxi_modules to load in.

Looking forward to your response. Thank you!

Best, Mika

guangkaixu commented 2 years ago

@mikacuy Hi, it seems that the trained weights of "auxi_modules" are not in our released checkpoint. If you would like to finetune our depth model on a special dataset, the auxi branch can be removed. In my opinion, the auxi branch are helpful to learn the output depth from disparity features during training, but can be not so significant during fine-tuning.

For the training code of "focal_keys" and "shift_keys", we will reorganize them but it may be time spending. Thank you for your following!