We provide the models trained by our methods.
(For the baseline model, we use the models which is pre-trained on VOC2007/2012-trainval provided by ChainerCV.)
Mean average precision (mAP) [%] for each method is shown below.
| clipart | watercolor | comic | |
|---|---|---|---|
| Baseline | 26.8 | 49.6 | 24.9 |
| DT | 38.0 | 50.4 | 29.8 |
| Proposed (DT+PL) | 46.0 | 54.3 | 37.2 |
Please download from here.