Amodal Completion on Pix2Gestalt ground-truth benchmark (test)
0.15GT-LPIPSCollaborative Multi-Agent Reasoning Framework
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Collaborative Multi-Agent Reasoning Framework2025.12 | 0.15 | 0.858 | 0.885 | 19.034 | 74.8 | 0.113 | 0.894 | 0.934 | 63.646 | 7.489 | |
| Pix2Gestalt2025.12 | 0.197 | 0.812 | 0.843 | 17.453 | 71.2 | 0.095 | 0.906 | 0.91 | 44.326 | 7.279 | |
| PD-MCEvaluation subset=Full Set2025.12 | 0.225 | 0.781 | 0.859 | 16.518 | 60.8 | 0.06 | 0.95 | 0.96 | 21.595 | 5.732 | |
| PD-MCEvaluation subset=Valid Only2025.12 | 0.279 | 0.751 | 0.817 | 16.22 | 53.7 | 0.216 | 0.82 | 0.857 | 16.578 | 5.333 | |
| OWAAC2025.12 | 0.311 | 0.784 | 0.757 | 13.642 | 43.6 | 0.194 | 0.8 | 0.871 | 25.658 | 5.833 |