Sketch-to-Photo Retrieval on QMUL-Shoes
69.6Recall@1IHDA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| IHDAModel configuration=Instance-level heterogeneous DA w/ att, w/ external Data=UT-Zap50K2022.11 | 69.6 | 97.4 | 99.1 | |
| IHDAw/ external Data=UT-Zap50K, Training/Evaluation Protocol=Instance-level heterogeneous DA w/ att2022.11 | 68.7 | 95.7 | — | |
| DSSA TripletModel configuration=pre-train w/o att + fine-tune, w/ external Data=TU Berlin Sketch & Edge-style ImageNet2022.11 | 61.9 | 95.1 | 98.6 | |
| DSSA Tripletw/ external Data=TU Berlin Sketch & Edge-style ImageNet, Training/Evaluation Protocol=pre-train w/o att + fine-tune2022.11 | 61.7 | 94.8 | — | |
| CD-AFLw/ external Data=TU Berlin Sketch & Edge-style ImageNet, Training/Evaluation Protocol=pre-train w/o att + fine-tune2022.11 | 56.4 | 92.6 | — | |
| MAR-FBIRw/ external Data=TU Berlin Sketch & Edge-style ImageNet, Training/Evaluation Protocol=pre-train w/o att + fine-tune2022.11 | 50.4 | 91.3 | — | |
| Triplet SNw/ external Data=TU Berlin Sketch & Edge-style ImageNet, Training/Evaluation Protocol=pre-train w/o att + fine-tune2022.11 | 39.1 | 87.8 | — |