search

Caoang327 / HexPlane

Official code for CVPR 2023 Paper, HexPlane: A Fast Representation for Dynamic Scenes
MIT License
242 stars 24 forks source link

Getting white images while taking ndc_rays=True #20

Open ayushi2610 opened 3 months ago

ayushi2610 commented 3 months ago

mask_filtered = torch.cat(mask_filtered).view(all_rgbs.shape[:-1]) RuntimeError: shape '[13080000]' is invalid for input of size 17440000 This is the error I am getting and if I take ndc_rays=True, the model works fine but I am getting plain white images in test and validation. What could be the problem?

Caoang327 commented 2 months ago

It looks like the torch.cat(mask_filtered) has different shape as all_rgbs.shape[:-1]. Which data are you using? Can you please double check that?

ayushi2610 commented 2 months ago

Hi, I am using DaVinci robotic surgery endoscopy video dataset.

On Fri, Apr 19, 2024, 11:10 PM Caoang327 @.***> wrote:

It looks like the torch.cat(mask_filtered) has different shape as all_rgbs.shape[:-1]. Which data are you using? Can you please double check that?

— Reply to this email directly, view it on GitHub https://github.com/Caoang327/HexPlane/issues/20#issuecomment-2067023768, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHORSVBLRDH3QVPVNXU5LO3Y6FJKPAVCNFSM6AAAAABFMRUQ2SVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANRXGAZDGNZWHA . You are receiving this because you authored the thread.Message ID: @.***>

Caoang327 commented 1 month ago

Sorry, I am not familiar with that dataset. But in general, you may want to check all_rays shape and all_rgbs shape. Their shapes' first two dimension should be the same