Object Detection on VOC
83.2AP@50CoKe
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CoKeMulti-crop training trick=false2021.05 | 83.2 | — | — | — | |
| MoCo-v2Multi-crop training trick=false2021.05 | 83 | — | — | — | |
| BYOLMulti-crop training trick=false2021.05 | 82.9 | — | — | — | |
| SwAVMulti-crop training trick=true2021.05 | 82.1 | — | — | — | |
| DINOMulti-crop training trick=true2021.05 | 82 | — | — | — | |
| Barlow TwinsMulti-crop training trick=false2021.05 | 81.5 | — | — | — | |
| SupervisedMulti-crop training trick=false2021.05 | 81.3 | — | — | — | |
| PyramidCLIPBackbone=ResNet50, Pre-train Dataset=143M2022.04 | 78.7 | 50.9 | 54.7 | — | |
| DECLIPBackbone=ResNet50, Pre-train Dataset=88M2022.04 | 77.4 | 50 | 53.6 | — | |
| CLIPBackbone=ResNet50, Pre-train Dataset=400M2022.04 | 73.5 | 45.5 | 47.9 | — | |
| RandomBackbone=ResNet502022.04 | 51.6 | 26.5 | 22.8 | — | |
| CutLERBackbone=ResNet-50, Zero-shot=true, Unsupervised=true, Class-agnostic=true2023.01 | 36.9 | — | — | 44.3 | |
| FreeSOLOBackbone=ResNet-101, Zero-shot=true, Unsupervised=true, Class-agnostic=true2023.01 | 15.9 | — | — | 21.3 | |
| InsDisProtocol=Object Det., Pre-training=ImageNet2021.12 | — | 48.82 | — | — | |
| MMCLProtocol=Object Det., Pre-training=ImageNet2021.12 | — | 50.73 | — | — | |
| MoCHiProtocol=Object Det., Pre-training=ImageNet2021.12 | — | 52.61 | — | — | |
| MoCoProtocol=Object Det., Pre-training=ImageNet2021.12 | — | 50.51 | — | — | |
| MoCo-v2Protocol=Object Det., Pre-training=ImageNet2021.12 | — | 50.67 | — | — | |
| PCL-v1Protocol=Object Det., Pre-training=ImageNet2021.12 | — | 53.93 | — | — | |
| PCL-v2Protocol=Object Det., Pre-training=ImageNet2021.12 | — | 53.92 | — | — | |
| SupervisedProtocol=Object Det., Pre-training=ImageNet2021.12 | — | 53.26 | — | — |