Research Article
Big Transfer Learning for Fine Art Classification
Table 4
State-of-the-art results for artist, genre, and style categorization on the original MultitaskPainting100k dataset and the resized one.
| Reference | Method | MultitaskPainting100k | Artist | Style | Genre | Average |
| [13] | STN + HOG | 56.5 | 57.2 | 63.6 | 56.85 | [34] | Structure selection | 65.50 | 63.15 | 67.83 | 65.49 | BiT-S (ours) | Big transfer-256 | 51.38 | 60.16 | 70.23 | 55.77 | BiT-M (ours) | Big transfer-256 | 66.04 | 64.32 | 71.97 | 65.18 | BiT-S (ours) | Big transfer-ori | 53.72 | 61.12 | 70.48 | 61.77 | BiT-M (ours) | Big transfer-ori | 68.14 | 65.82 | 72.64 | 68.87 |
|
|