| Q2L-CvT(ImageNet-21K pretrained, resolution 384) | 97.3 | Query2Label: A Simple Transformer Way to Multi-Label Classification | - |
| FeV+LV (pretrain from ImageNet) | 92.0 | Exploit Bounding Box Annotations for Multi-label Object Recognition | - |
| ViT-B-16 (ImageNet-21K pretrained) | 93.1 | ImageNet-21K Pretraining for the Masses | - |
| TResNet-L (resolution 448, pretrain from ImageNet) | 94.6 | Asymmetric Loss For Multi-Label Classification | - |
| ML-GCN (pretrain from ImageNet) | 94.0 | Multi-Label Image Recognition with Graph Convolutional Networks | - |
| Q2L-TResL(ImageNet-21K pretrained, resolution 448) | 96.9 | Query2Label: A Simple Transformer Way to Multi-Label Classification | - |
| MLD-TResNetL-AAM (resolution 448, pretrain from OpenImages V6) | 96.70 | Combining Metric Learning and Attention Heads For Accurate and Efficient Multilabel Image Classification | - |
| TResNet-L (resolution 448, pretrain from MS-COCO) | 95.8 | Asymmetric Loss For Multi-Label Classification | - |