We provide a collection models trained with ASL on various multi-label datasets
| Backbone | Input Size | Dataset | mAP |
|---|---|---|---|
| TResNet_M (ImageNet21K pretrain) | 224 | MS-COCO | 81.8 |
| TResNet_L | 448 | MS-COCO | 86.6 |
| TResNet_XL | 640 | MS-COCO | 88.4 |
| TResNet_L | 448 | OpenImagesV6 | 86.3 |
| TResNet_XL | 448 | PASCAL-VOC | 96.0 |
| TResNet_L | 448 | NUS-WIDE dataset split | 65.2 |