Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Improve training speed by 30%~40% #7

Merged
merged 8 commits into from
May 17, 2024
Merged

Improve training speed by 30%~40% #7

merged 8 commits into from
May 17, 2024

Conversation

hbb1
Copy link
Owner

@hbb1 hbb1 commented May 16, 2024

Fusing some operators, lead to better training speed. Note that all hyper-parameters and setting are kept the same.

Test on the whole DTU dataset with one RTX3090.

CD Time
3DGS 1.96 11.2min
2DGS (paper) 0.80 18.2min
2DGS (This) 0.79 10.9min

Tested on a NeRF Synthetic Chair with one RTX3090

EXP PSNR Time
2DGS (old) 35.35 12 min
2DGS (old T_precomp) 35.35 21 min
2DGS (This) 35.20 7 min 46s
2DGS (This T_precomp) 35.35 14min 50s

Tested on MipNeRF360 garden scene

EXP PSNR Time
2DGS (old) 26.89 50 min 56s
2DGS (This) 26.84 31 min 02s

@hbb1 hbb1 merged commit a97d67a into main May 17, 2024
@hbb1 hbb1 deleted the optim branch May 17, 2024 11:55
@hbb1 hbb1 restored the optim branch May 17, 2024 11:55
@hbb1 hbb1 deleted the optim branch May 17, 2024 18:34
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant