PVT v2: Improved Baselines with Pyramid Vision Transformer

Open paper
Benchmark Model Rank Results
image-classification-on-imagenetPVTv2-B4#376Top 1 Accuracy: 83.8%Number of params: 82MGFLOPs: 11.8
image-classification-on-imagenetPVTv2-B3#439Top 1 Accuracy: 83.2%Number of params: 45.2MGFLOPs: 6.9
image-classification-on-imagenetPVTv2-B2#565Top 1 Accuracy: 82%Number of params: 25.4MGFLOPs: 4
image-classification-on-imagenetPVTv2-B1#766Top 1 Accuracy: 78.7%Number of params: 13.1MGFLOPs: 2.1
image-classification-on-imagenetPVTv2-B0#956Top 1 Accuracy: 70.5%Number of params: 3.4MGFLOPs: 0.6
object-detection-on-coco-minivalSparse R-CNN (PVTv2-B2)#80box AP: 50.1AP50: 69.5AP75: 54.9
object-detection-on-coco-oPVTv2-B5 (Mask R-CNN)#23Average mAP: 28.2Effective Robustness: 6.85

Rank counts only results with a code link.