Metric: mAP (higher is better)
| # | Model↕ | mAP▼ | Augmentations | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | Q2L-CvT(ImageNet-21K pretrained, resolution 384) | 97.3 | No | Query2Label: A Simple Transformer Way to Multi-L... | 2021-07-22 | Code |
| 2 | Q2L-TResL(ImageNet-21K pretrained, resolution 448) | 96.9 | No | Query2Label: A Simple Transformer Way to Multi-L... | 2021-07-22 | Code |
| 3 | GKGNet | 96.8 | No | GKGNet: Group K-Nearest Neighbor based Graph Con... | 2023-08-28 | Code |
| 4 | MLD-TResNetL-AAM (resolution 448, pretrain from OpenImages V6) | 96.7 | No | Combining Metric Learning and Attention Heads Fo... | 2022-09-14 | Code |
| 5 | M3TR(448×448) | 96.5 | No | - | - | Code |
| 6 | Q2L-TResL(resolution 448) | 96.1 | No | Query2Label: A Simple Transformer Way to Multi-L... | 2021-07-22 | Code |
| 7 | MSRN(pretrain from MS-COCO) | 96 | No | Multi-layered Semantic Representation Network fo... | 2021-06-22 | Code |
| 8 | TResNet-L (resolution 448, pretrain from MS-COCO) | 95.8 | No | Asymmetric Loss For Multi-Label Classification | 2020-09-29 | Code |
| 9 | SSGRL (pretrain from MS-COCO) | 95 | No | Learning Semantic-Specific Graph Representation ... | 2019-08-20 | Code |
| 10 | TDRG-R101(448×448) | 95 | No | Transformer-based Dual Relation Graph for Multi-... | 2021-10-10 | Code |
| 11 | MCAR (ResNet101, 448x448) | 94.8 | No | Learning to Discover Multi-Class Attentional Reg... | 2020-07-03 | Code |
| 12 | TResNet-L (resolution 448, pretrain from ImageNet) | 94.6 | No | Asymmetric Loss For Multi-Label Classification | 2020-09-29 | Code |
| 13 | ML-GCN (pretrain from ImageNet) | 94 | No | Multi-Label Image Recognition with Graph Convolu... | 2019-04-07 | Code |
| 14 | SSGRL (pretrain from ImageNet) | 93.4 | No | Learning Semantic-Specific Graph Representation ... | 2019-08-20 | Code |
| 15 | Ours PF-DLDL | 93.4 | No | Deep Label Distribution Learning with Label Ambi... | 2016-11-06 | Code |
| 16 | ViT-B-16 (ImageNet-21K pretrained) | 93.1 | No | ImageNet-21K Pretraining for the Masses | 2021-04-22 | Code |
| 17 | FeV+LV (pretrain from ImageNet) | 92 | No | Exploit Bounding Box Annotations for Multi-label... | 2015-04-22 | - |