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
652 stars 62 forks source link

Update test_sagemaker.yaml #44

Closed wanmeihuali closed 1 year ago

github-actions[bot] commented 1 year ago

Running experiment on sagemaker with git sha 2fc8adfab0113ea5f2b39a9fe2baca3ef0f4fe05

github-actions[bot] commented 1 year ago

Training job tune-param-reset-alpha-2fc8adf-230624-071507-tat-truck created

github-actions[bot] commented 1 year ago

Training job tune-param-reset-alpha-2fc8adf-230624-071507-tat-truck completed.

Model url: s3://taichi-3d-gaussian-splatting-log/tat-truck/tune-param-reset-alpha-2fc8adf-230624-071507-tat-truck/output/model.tar.gz,

tensorboard output path: s3://taichi-3d-gaussian-splatting-log/tat-truck/tune-param-reset-alpha-2fc8adf-230624-071507-tat-truck/output/output.tar.gz

github-actions[bot] commented 1 year ago

Training job tune-param-reset-alpha-2fc8adf-230624-071507-tat-truck final metrics:

Latest Metrics

train:psnr train:num_valid_points val:ssim train:loss train:ssim train:ssimloss val:loss val:psnr train:iteration train:l1loss
25.726137161254883 310056.0 0.7951681017875671 0.06562433391809464 0.7947728037834167 0.20522719621658325 0.07889925688505173 22.37520980834961 30000.0 0.03072362020611763

Max Metrics

train:psnr train:7kpsnr val:ssim train:ssim val:5kpsnr val:7kpsnr val:psnr train:5kssim train:5kpsnr val:7kssim val:5kssim train:7kssim
25.726137161254883 23.59933853149414 0.7951681017875671 0.7947728037834167 22.236431121826172 22.817672729492188 22.37520980834961 0.770331859588623 22.242815017700195 0.7887746095657349 0.7727715373039246 0.7896139621734619