3D fur geometry reconstruction on Synthetic Tiger Asset (test)
26.22Precision (2/20)NeuralFur
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| NeuralFurmode=proposed2026.01 | 26.22 | 39.32 | 48.05 | 20.58 | 34.08 | 45.69 | 23.06 | 36.51 | 46.84 | |
| NeuralFur (w/ fix. length)ablation=fixed length modeling2026.01 | 17.18 | 26.65 | 33.73 | 22.24 | 35.08 | 45.44 | 19.38 | 30.29 | 38.72 | |
| Gaussian Haircutstage=2nd stage2026.01 | 16.24 | 25.51 | 32.34 | 23.51 | 36.04 | 45.87 | 19.21 | 29.87 | 37.93 | |
| Gaussian Haircutstage=3rd stage2026.01 | 5.78 | 10.44 | 15.43 | 21.39 | 32.82 | 42.36 | 9.1 | 15.84 | 22.62 | |
| NeuralFur (w/ Unet)ablation=U-Net feature extraction2026.01 | 5.77 | 12.16 | 20.06 | 7.05 | 15.39 | 25.8 | 6.35 | 13.58 | 22.57 | |
| NeuralFur (w/ opt.length)ablation=optimized length modeling2026.01 | 3.05 | 6.26 | 9.82 | 16.63 | 28.27 | 38.45 | 5.16 | 10.25 | 15.65 |