R
ReflectIris
← Back to Training Runs

exp_161_clean_ml02

Completeml-02

Started

29.3 dB
Train PSNR
Momentum
Plateau
Avg PSNR gaining slowly (+-0.05 dB / 5K steps)
PSNR Δ5K: -0.05 dB
Gaussian growth: +179,982
Loss trend: Increasing
Visual Analysis
STALL
Visual quality plateaued (SSIM range 0.0082 over last 3 checkpoints) (31 checkpoints analyzed)
SSIM
0.896
Blur Ratio
0.233
Edge Preservation
0.572
Expression Fidelity
1.000
Progress
0%
60,000 / 0 steps
Gaussians
200.0K
Growing
Speed
Duration
Densify Window
0K–0K
0 events
Batch Size
1
Step 60,0000 total
densify

Training Metrics

Train + Val PSNR (dB) — Higher is Better

Training Loss — Lower is Better

Gaussian Count — Grows During Densification

Training Speed (iterations/sec) & GPU Memory (GB)

Visual Progress

Step 60Kml-02
·
Step 58Kml-02
·
Step 56Kml-02
·
Step 54Kml-02
·
Step 52Kml-02
·
Step 50Kml-02
·
Step 48Kml-02
·
Step 46Kml-02
·
Step 44Kml-02
·
Step 42Kml-02
·
Step 40Kml-02
·
Step 38Kml-02
·
Step 36Kml-02
·
Step 34Kml-02
·
Step 32Kml-02
·
Step 30Kml-02
·
Step 28Kml-02
·
Step 26Kml-02
·
Step 24Kml-02
·
Step 22Kml-02
·
Step 20Kml-02
·
Step 18Kml-02
·
Step 16Kml-02
·
Step 14Kml-02
·
Step 12Kml-02
·
Step 10Kml-02
·
Step 8Kml-02
·
Step 6Kml-02
·
Step 4Kml-02
·
Step 2Kml-02
·
Step 1ml-02
·
Full Configuration
{}