Research Article
A New Algorithm for Sketch-Based Fashion Image Retrieval Based on Cross-Domain Transformation
Table 4
Accuracy comparison with baselines on the Fashion Image dataset, QMUL-shoes and QMUL-chairs datasets.
| | Methods | Fashion Image dataset | QMUL-shoes | QMUL-chairs | | acc.@1 | acc.@10 | acc.@1 | acc.@10 | acc.@1 | acc.@10 |
| | BoW-HOG + rank-SVM [6] | — | — | 0.174 | 0.678 | 0.289 | 0.670 | | ISN Deep + rank-SVM [37] | — | — | 0.200 | 0.626 | 0.474 | 0.825 | | Dense-HOG + rank-SVM [8] | — | — | 0.244 | 0.652 | — | — | | 3DS Deep + rank-SVM [10] | — | — | 0.052 | 0.217 | 0.062 | 0.268 | | Sketchy [17] | 0.673 | 0.953 | — | — | — | — | | Our model | 0.924 | 0.977 | 0.308 | 0.654 | 0.495 | 0.794 |
|
|
