Data preprocessing needed to fit format

Tangshitao / MVDiffusion

MVDiffusion: Enabling Holistic Multi-view Image Generation with Correspondence-Aware Diffusion, NeurIPS 2023 (spotlight)

447 stars 21 forks source link

Data preprocessing needed to fit format #44

Closed jbu5105 closed 3 months ago

jbu5105 commented 4 months ago

Hi! Thank you very much for your work. I downloaded data from Scannet and the labels link from dropbox and I have the resulting structure per scene:

Captura de pantalla 2024-03-07 173113

However in the README you specify the following format.

Even by converting the instance to color, in the dataloader it expects a .jpg and not .png. And still I would be missing depth, pose, intrinsic... Is there any preprocessing step that I am missing?

jbu5105 commented 3 months ago

Solved in other issue: https://github.com/Tangshitao/MVDiffusion/issues/21