← Run Index
Run summary · P1 Phase G.5 final Pareto integration
chair-phase-i-adaptive-12500
Phase I scene-adaptive: chair @ stop_iter=12500 (fast converger 仮説検証) variant (early stop + SH progressive)
partial pass
01
設定
- dataset
- /Users/otkrickey/dev/3dgs-workspace/datasets/nerf_synthetic/chair
- 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
35.88dB
Wallclock
18m 35s
Final splats
794,562
Final loss
4.628078e-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 | 6.0968e-1 |
| 1,000 | 5.0307e-2 |
| 2,000 | 1.2539e-2 |
| 3,000 | 7.9335e-3 |
| 4,000 | 6.6839e-3 |
| 5,000 | 5.9937e-3 |
| 6,000 | 5.6577e-3 |
| 7,000 | 5.3519e-3 |
| 8,000 | 5.1740e-3 |
| 9,000 | 4.9816e-3 |
| 10,000 | 4.8257e-3 |
| 11,000 | 4.7950e-3 |
| 12,000 | 4.7039e-3 |
| 12,500 | 4.6281e-3 |
06
関連 finding
この run を related_runs に含む finding (Positive / Negative どちらも)。