search

wanmeihuali / taichi_3d_gaussian_splatting

An unofficial implementation of paper 3D Gaussian Splatting for Real-Time Radiance Field Rendering by taichi lang.
Apache License 2.0
659 stars 62 forks source link

Adjustable tile size #127

Closed wanmeihuali closed 1 year ago

wanmeihuali commented 1 year ago

Support adjustable tile size; name symbols to present misunderstanding; add nvprof files to gitignore;

github-actions[bot] commented 1 year ago

Running experiment on sagemaker with git sha 834b5fcfd86f5fab6d1a1a5012dd2d7ad8abffd4

github-actions[bot] commented 1 year ago

Training job adjustable-TILE-SIZE-834b5fc-230921-073413-tat-truck-baseline created

github-actions[bot] commented 1 year ago

Running experiment on sagemaker with git sha 834b5fcfd86f5fab6d1a1a5012dd2d7ad8abffd4

github-actions[bot] commented 1 year ago

Training job adjustable-TILE-SIZE-834b5fc-230921-073832-tat-train-baseline created

github-actions[bot] commented 1 year ago

Training job adjustable-TILE-SIZE-834b5fc-230921-073413-tat-truck-baseline completed.

Model url: s3://taichi-3d-gaussian-splatting-log/tat-truck-baseline/adjustable-TILE-SIZE-834b5fc-230921-073413-tat-truck-baseline/output/model.tar.gz,

tensorboard output path: s3://taichi-3d-gaussian-splatting-log/tat-truck-baseline/adjustable-TILE-SIZE-834b5fc-230921-073413-tat-truck-baseline/output/output.tar.gz

github-actions[bot] commented 1 year ago

Training job adjustable-TILE-SIZE-834b5fc-230921-073413-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 0.025658147409558296 0.03904088959097862 439899.0 27.63935089111328 0.9074281454086304 0.09257185459136963 0.05359480530023575 25.16960334777832 0.8661096692085266

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
24.174163818359375 0.8364933729171753 23.10714340209961 0.8057988882064819 27.63935089111328 0.9074281454086304 23.10798454284668 0.8160789012908936 23.73977279663086 0.8349094986915588 25.16960334777832 0.8661096692085266
github-actions[bot] commented 1 year ago

Training job adjustable-TILE-SIZE-834b5fc-230921-073832-tat-train-baseline completed.

Model url: s3://taichi-3d-gaussian-splatting-log/tat-train-baseline/adjustable-TILE-SIZE-834b5fc-230921-073832-tat-train-baseline/output/model.tar.gz,

tensorboard output path: s3://taichi-3d-gaussian-splatting-log/tat-train-baseline/adjustable-TILE-SIZE-834b5fc-230921-073832-tat-train-baseline/output/output.tar.gz

github-actions[bot] commented 1 year ago

Training job adjustable-TILE-SIZE-834b5fc-230921-073832-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 0.03872641548514366 0.05631912499666214 678651.0 25.311317443847656 0.8733100295066833 0.12668997049331665 0.09814545512199402 21.046857833862305 0.802087128162384

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
20.07990074157715 0.653599739074707 21.625051498413086 0.7544162273406982 25.311317443847656 0.8733100295066833 19.043289184570312 0.7301173806190491 19.716796875 0.7571879625320435 21.046857833862305 0.802087128162384