← Run Index
Run summary · P1 Phase G.5 final Pareto integration
lego-phase-h-pareto-12500
Phase H Pareto sweep @ stop_iter=12500 variant (early stop + SH progressive)
partial pass
01
設定
- dataset
- /Users/otkrickey/dev/3dgs-workspace/datasets/nerf_synthetic/lego
- gt_convention
- Premultiplied
- iterations
- 12,500
- seed
- 42
- capacity
- 1,000,000 splats
- sh_degree
- 3
- loss
- L1Ssim
- lambda
- 0.200
- ssim
- window=7 sigma=1
- backend.backward
- Simd
- backend.loss_path
- Gpu
02
結果
Val PSNR
36.26dB
Wallclock
11m 48s
Final splats
407,053
Final loss
5.438210e-3
03
3D ビューワー
学習結果 PLY を .splat 形式に変換し、antimatter15/splat の WebGL ビューワーで対話表示。
04
推移
Loss curve — 反復ごとの training loss
Splat count — refine の発火と最終構成
05
中間ログ抜粋
| iter | 本 run loss |
|---|---|
| 1 | 5.9384e-1 |
| 1,000 | 5.4218e-2 |
| 2,000 | 1.4075e-2 |
| 3,000 | 8.4029e-3 |
| 4,000 | 7.3230e-3 |
| 5,000 | 6.5188e-3 |
| 6,000 | 6.2119e-3 |
| 7,000 | 5.9410e-3 |
| 8,000 | 5.8092e-3 |
| 9,000 | 5.7231e-3 |
| 10,000 | 5.6782e-3 |
| 11,000 | 5.6376e-3 |
| 12,000 | 5.5876e-3 |
| 12,500 | 5.4382e-3 |