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 binary input #90

Closed wanmeihuali closed 1 year ago

github-actions[bot] commented 1 year ago

Running experiment on sagemaker with git sha 6c6e0c278bee376588de7b22960fd15b78b0d68b

github-actions[bot] commented 1 year ago

Training job support-binary-input-6c6e0c2-230711-030719-tat-truck created

github-actions[bot] commented 1 year ago

Training job support-binary-input-6c6e0c2-230711-030719-tat-truck completed.

Model url: s3://taichi-3d-gaussian-splatting-log/tat-truck/support-binary-input-6c6e0c2-230711-030719-tat-truck/output/model.tar.gz,

tensorboard output path: s3://taichi-3d-gaussian-splatting-log/tat-truck/support-binary-input-6c6e0c2-230711-030719-tat-truck/output/output.tar.gz

github-actions[bot] commented 1 year ago

Training job support-binary-input-6c6e0c2-230711-030719-tat-truck final metrics:

Latest Metrics

train:num_valid_points train:ssimloss train:psnr train:loss val:loss val:ssim train:l1loss train:iteration val:psnr train:ssim
436488.0 0.153387188911438 25.021038055419922 0.05578380823135376 0.05988791957497597 0.8409193158149719 0.031382959336042404 30000.0 24.584850311279297 0.846612811088562

Max Metrics

val:7kssim train:psnr train:7kpsnr val:7kpsnr val:ssim train:7kssim val:5kpsnr train:5kssim val:psnr train:ssim train:5kpsnr val:5kssim
0.8073482513427734 25.021038055419922 23.105337142944336 23.251049041748047 0.8409193158149719 0.7839511632919312 22.57291030883789 0.7839922308921814 24.584850311279297 0.846612811088562 22.16071319580078 0.7846093773841858