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.

Sign up for GitHub

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

Jump to bottom

Enable Pose Optimization #136

Merged
merged 35 commits into from
Oct 6, 2023
Merged

Enable Pose Optimization #136

merged 35 commits into from
Oct 6, 2023

Conversation

wanmeihuali
Copy link
Owner

No description provided.

@wanmeihuali
move normalization outside, and use precompute activation
059e356
@wanmeihuali
further optimize structure.
52cea90
@wanmeihuali
remove useless variables.
967c8e3
@wanmeihuali
draft for grad pose
54cde10
@wanmeihuali
Merge branch 'main' into grad_pose
c031afb
@wanmeihuali
update d_uv compute function.
93dd861
@wanmeihuali
reduce sum for grad.
ada3fd6
@wanmeihuali
fix bug.
a45495d
@wanmeihuali
adapt dataset for camera pose optimization.
15e0420
@wanmeihuali
enable pose optimize
c5dd5ca
@wanmeihuali
enable save optimized pose
8403daf
@wanmeihuali
fix bug: crash when model dir not exist.
7f847c6
@wanmeihuali
add script for pose_estimation, it shows pose grad works.
83878c5
@wanmeihuali
Merge branch 'main' into grad_pose
7ad3677
@wanmeihuali
try optimize with pose refinement
b5d19e7
@wanmeihuali
Merge branch 'grad_pose' of https://github.com/wanmeihuali/taichi_3d_…
ece92b0 
…gaussian_splatting into grad_pose
@wanmeihuali
use SGD for pose shall be a better idea. Or we shall make gradient sp…
02697e4 
…arse.
@wanmeihuali
add decay for camera pose
71922f2
@wanmeihuali
Merge branch 'main' into grad_pose
d187e3b
@wanmeihuali
disable camera pose optimization
2359a4d
@wanmeihuali
Merge branch 'main' into grad_pose
186d437
@wanmeihuali
fix running error.
ee644c9
@wanmeihuali
before reverse input data
03842b1
@wanmeihuali
get pose estimation work.
0f644c1
@wanmeihuali
Merge branch 'main' into inverse_saved_pose
e0fe458
@wanmeihuali wanmeihuali changed the title Inverse saved pose Enable Pose Optimization Oct 5, 2023
@github-actions
Copy link

github-actions bot commented Oct 5, 2023

Running experiment on sagemaker with git sha fbbdaf4

@github-actions
Copy link

github-actions bot commented Oct 5, 2023

Training job inverse-saved-pose-fbbdaf4-231005-041642-tat-train-baseline created

@github-actions
Copy link

github-actions bot commented Oct 5, 2023

Running experiment on sagemaker with git sha 4635463

@github-actions
Copy link

github-actions bot commented Oct 5, 2023

Training job inverse-saved-pose-4635463-231005-225925-tat-train-baseline created

@github-actions
Copy link

github-actions bot commented Oct 5, 2023

Running experiment on sagemaker with git sha 4635463

@github-actions
Copy link

github-actions bot commented Oct 5, 2023

Training job inverse-saved-pose-4635463-231005-230050-tat-truck-baseline created

@github-actions
Copy link

github-actions bot commented Oct 6, 2023

Training job inverse-saved-pose-4635463-231005-230050-tat-truck-baseline completed.

Model url: s3://taichi-3d-gaussian-splatting-log/tat-truck-baseline/inverse-saved-pose-4635463-231005-230050-tat-truck-baseline/output/model.tar.gz,

tensorboard output path: s3://taichi-3d-gaussian-splatting-log/tat-truck-baseline/inverse-saved-pose-4635463-231005-230050-tat-truck-baseline/output/output.tar.gz

@github-actions
Copy link

github-actions bot commented Oct 6, 2023

Training job inverse-saved-pose-4635463-231005-230050-tat-truck-baseline final metrics:

Latest Metrics

train:iteration train:l1loss train:loss train:num_valid_points train:psnr train:ssim train:ssimloss val:loss val:psnr val:ssim
30000.0 0.03413218632340431 0.056418053805828094 442597.0 24.21188735961914 0.8544384837150574 0.14556151628494263 0.05384311452507973 25.154382705688477 0.865390419960022

Max Metrics

train:5kpsnr train:5kssim train:7kpsnr train:7kssim train:psnr train:ssim val:5kpsnr val:5kssim val:7kpsnr val:7kssim val:psnr val:ssim
24.060989379882812 0.8435416221618652 25.38058090209961 0.8602480888366699 24.21188735961914 0.8544384837150574 23.16160011291504 0.8166438937187195 23.83233070373535 0.8370955586433411 25.154382705688477 0.865390419960022

@github-actions
Copy link

github-actions bot commented Oct 6, 2023

Training job inverse-saved-pose-4635463-231005-225925-tat-train-baseline completed.

Model url: s3://taichi-3d-gaussian-splatting-log/tat-train-baseline/inverse-saved-pose-4635463-231005-225925-tat-train-baseline/output/model.tar.gz,

tensorboard output path: s3://taichi-3d-gaussian-splatting-log/tat-train-baseline/inverse-saved-pose-4635463-231005-225925-tat-train-baseline/output/output.tar.gz

@github-actions
Copy link

github-actions bot commented Oct 6, 2023

Training job inverse-saved-pose-4635463-231005-225925-tat-train-baseline final metrics:

Latest Metrics

train:iteration train:l1loss train:loss train:num_valid_points train:psnr train:ssim train:ssimloss val:loss val:psnr val:ssim
30000.0 0.0653761699795723 0.08833681046962738 669415.0 20.82976531982422 0.8198206424713135 0.18017935752868652 0.09855514764785767 21.113161087036133 0.8022959232330322

Max Metrics

train:5kpsnr train:5kssim train:7kpsnr train:7kssim train:psnr train:ssim val:5kpsnr val:5kssim val:7kpsnr val:7kssim val:psnr val:ssim
19.03648567199707 0.654522716999054 20.009803771972656 0.8380850553512573 20.82976531982422 0.8198206424713135 19.09685516357422 0.7308458089828491 19.79014778137207 0.7569055557250977 21.113161087036133 0.8022959232330322

@wanmeihuali wanmeihuali merged commit 018c1e8 into main Oct 6, 2023
1 check passed
@wanmeihuali wanmeihuali deleted the inverse_saved_pose branch October 6, 2023 01:18
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@yanzhoupan yanzhoupan Awaiting requested review from yanzhoupan

Assignees
No one assigned
Labels
need_experiment_tat_train_baseline need_experiment_tat_truck_baseline
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

None yet
1 participant
@wanmeihuali