Object Detection on PASCAL VOC (AP50, AP, AP75)
82.4AP50MoCo-v2
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MoCo-v2Pretrain Dataset=IN-1K, Pretrain Epochs=200, Backbone=ResNet-50, Fine-tuning=true2021.04 | 82.4 | 57 | 63.6 | |
| Multimodal Contrastive Training (scratch)Pretrain Dataset=COCO, Pretrain Epochs=1000, Backbone=ResNet-50, Fine-tuning=true2021.04 | 82.1 | 56.1 | 62.4 | |
| Multimodal Contrastive Training (with tag)Pretrain Dataset=COCO, Pretrain Epochs=200, Backbone=ResNet-50, Fine-tuning=true2021.04 | 81.8 | 55.8 | 61.7 | |
| IN-SupPretrain Dataset=IN-1K, Pretrain Epochs=90, Backbone=ResNet-50, Fine-tuning=true2021.04 | 81.6 | 54.3 | 59.7 | |
| MoCoPretrain Dataset=IN-1K, Pretrain Epochs=200, Backbone=ResNet-50, Fine-tuning=true2021.04 | 81.5 | 55.9 | 62.6 | |
| VirTexPretrain Dataset=COCO, Pretrain Epochs=1000, Backbone=ResNet-50, Fine-tuning=true2021.04 | 81.4 | 55.6 | 61.5 | |
| Multimodal Contrastive TrainingPretrain Dataset=COCO, Pretrain Epochs=200, Backbone=ResNet-50, Fine-tuning=true2021.04 | 80.8 | 55.6 | 61.9 | |
| Multimodal Contrastive Training (scratch)Pretrain Dataset=COCO, Pretrain Epochs=200, Backbone=ResNet-50, Fine-tuning=true2021.04 | 80.7 | 55.4 | 61.5 | |
| VirTex*Pretrain Dataset=COCO, Pretrain Epochs=200, Backbone=ResNet-50, Fine-tuning=true2021.04 | 80.2 | 54.8 | 60.9 | |
| MoCo-v2Pretrain Dataset=COCO, Pretrain Epochs=200, Backbone=ResNet-50, Fine-tuning=true2021.04 | 75.5 | 48.4 | 52.1 | |
| MoCoPretrain Dataset=COCO, Pretrain Epochs=200, Backbone=ResNet-50, Fine-tuning=true2021.04 | 75.4 | 47.5 | 51.1 | |
| Random InitPretrain Dataset=NA, Pretrain Epochs=NA, Backbone=ResNet-50, Fine-tuning=true2021.04 | 60.2 | 33.8 | 33.1 |