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
666 stars 62 forks source link

precompute point color #50

Closed wanmeihuali closed 1 year ago

github-actions[bot] commented 1 year ago

Running experiment on sagemaker with git sha 851e66071f27f706fbc09625342a16b2419bcb04

github-actions[bot] commented 1 year ago

Training job engineering-precompute-color-851e660-230625-042907-tat-truck created

github-actions[bot] commented 1 year ago

Training job engineering-precompute-color-851e660-230625-042907-tat-truck completed.

Model url: s3://taichi-3d-gaussian-splatting-log/tat-truck/engineering-precompute-color-851e660-230625-042907-tat-truck/output/model.tar.gz,

tensorboard output path: s3://taichi-3d-gaussian-splatting-log/tat-truck/engineering-precompute-color-851e660-230625-042907-tat-truck/output/output.tar.gz

github-actions[bot] commented 1 year ago

Training job engineering-precompute-color-851e660-230625-042907-tat-truck final metrics:

Latest Metrics

train:ssimloss train:iteration val:loss val:ssim train:psnr train:ssim train:loss train:num_valid_points train:l1loss val:psnr
0.12455093860626221 30000.0 0.0606846958398819 0.8400782942771912 26.53762435913086 0.8754490613937378 0.045441098511219025 429656.0 0.02566363662481308 24.445302963256836

Max Metrics

train:5kpsnr val:ssim val:7kpsnr train:7kssim train:psnr val:5kssim val:5kpsnr train:ssim train:7kpsnr train:5kssim val:7kssim val:psnr
23.55860710144043 0.8400782942771912 23.018129348754883 0.831803560256958 26.53762435913086 0.7760510444641113 22.284759521484375 0.8754490613937378 23.879531860351562 0.7924162745475769 0.8016843199729919 24.445302963256836