search

liuyuan-pal / SyncDreamer

[ICLR 2024 Spotlight] SyncDreamer: Generating Multiview-consistent Images from a Single-view Image
https://liuyuan-pal.github.io/SyncDreamer/
MIT License
906 stars 39 forks source link

picture from gso adaption #59

Open clear-train opened 9 months ago

clear-train commented 9 months ago

What did you do with the images in the GSO dataset to conform to the input format? I've noticed that the pictures in the gso dataset all seem to be 3-channel

liuyuan-pal commented 8 months ago

Hi. The rendered images have RGBA channels. Then, I will convert the background to white.