zhyever

That could be something related to 'https://github.com/zhyever/Monocular-Depth-Estimation-Toolbox/blob/main/depth/datasets/custom.py'. If you def a new CustomDepthDataset class, there could be such a error thrown out, because the CustomDepthDataset has already been registered in...

I was wondering if this process is out of memory? And it would be killed by the system.

Close for now.

Close it for now. Feel free to re-open for further discussion.

Close it for now.

Hope everything goes well! I close this issue for now. Feel free to re-open it.

Sorry for the late. Really busy with applications. The config is uploaded now. Close this issue for now.

Could you please explain your question more specifically? I wonder if you mean you resize the images/gt during the inference stage, but get inferior performance?

Hi, I'm sorry for this late reply. 1. The logic is that we first crop images to 352x1216 following the previous [KBCrop](https://github.com/zhyever/Monocular-Depth-Estimation-Toolbox/blob/7c0c89c8db07631fec1737f3087e4f1f540ccd53/depth/datasets/pipelines/transforms.py#L89). It is a center crop and the aim...

I think the raw resolution is better if there is no invalid GT value lying in the fringe of images. If we have to crop, I guess we can calculate...