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
648 stars 61 forks source link

support initial with color #88

Closed wanmeihuali closed 1 year ago

github-actions[bot] commented 1 year ago

Running experiment on sagemaker with git sha 64744b0d7021e04dc9f1bf75de4843cc36913085

github-actions[bot] commented 1 year ago

Training job initial-with-color-64744b0-230710-063008-tat-truck created

github-actions[bot] commented 1 year ago

Training job initial-with-color-64744b0-230710-063008-tat-truck completed.

Model url: s3://taichi-3d-gaussian-splatting-log/tat-truck/initial-with-color-64744b0-230710-063008-tat-truck/output/model.tar.gz,

tensorboard output path: s3://taichi-3d-gaussian-splatting-log/tat-truck/initial-with-color-64744b0-230710-063008-tat-truck/output/output.tar.gz

github-actions[bot] commented 1 year ago

Training job initial-with-color-64744b0-230710-063008-tat-truck final metrics:

Latest Metrics

train:iteration val:psnr train:psnr train:ssimloss val:ssim train:loss train:num_valid_points val:loss train:l1loss train:ssim
30000.0 24.474742889404297 25.90972900390625 0.17083287239074707 0.8399139642715454 0.05877495929598808 431054.0 0.06031208485364914 0.03076048009097576 0.8291671276092529

Max Metrics

train:5kssim val:psnr train:psnr val:5kpsnr train:7kpsnr val:ssim val:7kpsnr val:5kssim train:7kssim train:ssim val:7kssim train:5kpsnr
0.7833477258682251 24.474742889404297 25.90972900390625 22.311662673950195 23.742183685302734 0.8399139642715454 23.051403045654297 0.7784926295280457 0.8307317495346069 0.8291671276092529 0.8026293516159058 22.731000900268555