Centroid Transformers: Learning to Abstract with Attention
Open paper
Benchmark
Model
Rank
Results
image-classification-on-imagenet
CentroidViT-S (arXiv, 2021-02)
–
Top 1 Accuracy: 80.9%
Number of params: 22.3M
GFLOPs: 9.4
Rank counts only results with a code link.