which part of nyuv2

[Xiang-cd]

there are different source of nyuv2:

• https://huggingface.co/datasets/tanganke/nyuv2
• https://huggingface.co/datasets/sayakpaul/nyu_depth_v2
• https://cs.nyu.edu/~fergus/datasets/nyu_depth_v2.html

which one do you use, and do you have some scripts to process them into a standard format?

[Xiang-cd]

oh, i see, only use the test part to do evaluation..

[Xiang-cd]

but i still wondering how to do evaluation

[gaoxiangjun]

You may see the README and ./benchmark

[Xiang-cd]

I think I didn't find a readme file at ./benchmark dir?

[gaoxiangjun]

https://github.com/Tencent/DepthCrafter?tab=readme-ov-file#-dataset-evaluation

[Xiang-cd]

$ tree -L 2 130 ↵
.
├── download_from_gdrive.py
├── nyu_depth_v2_labeled.mat
├── sync
│ ├── basement_0001a
│ ├── basement_0001b
│ ├── bathroom_0001
│ ├── bathroom_0002
│ ├── bathroom_0005
│ ├── bathroom_0006
│ ├── bathroom_0007

thanks, I've readed the readme and the code already, but my downloaded nyuv2 file tree, looks like this, and I could find "path/to/NYUv2/filename_list_test.txt" in my downloaded version, how to align the file tree and file list text file with yours?

[Xiang-cd]

thank you for your great work~ wish a help

[gaoxiangjun]

filename_list_test.txt is in the dataset. filename_list_test.txt

You can also use this.

[zhizunhu]

@gaoxiangjun I just got test file from nyu_depth_v2_labeled.mat, RGB image name is same as your filename_list_test.txt，But I only get delta1=0.924，more worse than 0.944 in paper. my gpu is a100-80g. cc @Xiang-cd Have you reproduced nyuv2 metric? Wish your help, Thanks.

[wbhu]

You may refer to the released benchmark tools for evaluation details