Can you provide a specific explanation for this part of the description?

[Ricky-PLUS]

“Please download the Open Spatial Dataset from huggingface, and modify the path in llava/data/dataset_mixture.py.

For raw images, please download OpenImages from OpenImagesV7. To process the rgb images into depth, we use DepthAnythingV2 and save the depth with the following function:”

May I ask which part of the V7 website you provided to download from? And is your intention to process the downloaded image with the code you provided to generate a depth image?

[peiwang062]

The same question. How to render annotations from huggingface onto OpenImagesV7 images to obtain training images similar to Figure 2 in the paper?