Retrieval on CompCars
97.7Top-1 Instance AccuracyPoint-BERT(L)
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Point-BERT(L)Pre-train Data=Ensembled [26], Fine-tuned=multi-modal InfoNCE loss2026.03 | 97.7 | — | 100 | — | 98.8 | — | |
| Point-BERT(S)Pre-train Data=Shapenet [1], Fine-tuned=multi-modal InfoNCE loss2026.03 | 93.8 | — | 99.9 | — | 96.5 | — | |
| SparseConvPre-train Data=Shapenet [1], Fine-tuned=multi-modal InfoNCE loss2026.03 | 89.5 | — | 99.8 | — | 93.8 | — | |
| SparseConvPre-train Data=Ensembled [26], Fine-tuned=multi-modal InfoNCE loss2026.03 | 85.1 | — | 99.2 | — | 90.5 | — | |
| SC-IBSR2026.03 | 78.7 | — | 94.2 | — | — | — | |
| CMIC2026.03 | 77.8 | — | 94.1 | — | — | — | |
| HEG-TS2026.03 | 67.1 | — | 93.7 | — | — | — | |
| LFD2026.03 | 20.5 | — | 58 | — | — | — |