Sliced Recursive Transformer

Open paper
Benchmark Model Rank Results
image-classification-on-imagenetSReT-B (384 res, ImageNet-1K only)#279Top 1 Accuracy: 84.8%Number of params: 71.2M
image-classification-on-imagenetSReT-S (512 res, ImageNet-1K only)#319Top 1 Accuracy: 84.3%Number of params: 21.3MGFLOPs: 42.8
image-classification-on-imagenetSReT-S (384 res, ImageNet-1K only)#377Top 1 Accuracy: 83.8%Number of params: 21MGFLOPs: 18.5
image-classification-on-imagenetSReT-T#817Top 1 Accuracy: 77.6%Number of params: 4.8MGFLOPs: 1.1
image-classification-on-imagenetSReT-ExT#919Top 1 Accuracy: 74.0%Number of params: 4MGFLOPs: 0.7

Rank counts only results with a code link.