Data relationship difference between v1 and v2

[loveaca-sunlight]

Hi there, thanks a lot for the work on CO3D it helps me a lot.

Before that, I have done some experiments on the v1 data, and recently I want to update it to the v2 data version. But I found some differences in data naming between the two.

I noticed that one category of data is classified as "test_known"/"test_unseen"/"train_known"/"train_unseen" in v1.

However, in v2, there are "dev"/"train"/"test" categories, and there are still "train"/"test" in its subordinate categories. After reading the description of v2 in the readme document, I am still a little confused.

Could you please help me explain the difference and differences from v1 to v2? If possible, I hope to give some guidance so that I can use the data classification of the v1 version to process the v2 data. Thank you very much for your help.

[davnov134]

Hi,

there's a quite verbose answer in one of the previous issues: https://github.com/facebookresearch/co3d/issues/49#issuecomment-1203678523

Let us know if this helps.

[zhaomingheyuhan]

Hello! I am confused about 'known' and 'unseen'.Does 'known' representative know the camera parameters?