Learning Correlation Structures for Vision Transformers

Benchmark Model Rank Results
image-classification-on-imagenetStructViT-L-4-1* up384#139Top 1 Accuracy: 86.7%
action-classification-on-kinetics-400StructViT-B-4-1–Acc@1: 83.4
action-recognition-in-videos-on-something-something-v1StructVit-B-4-1–Top 1 Accuracy: 61.3
action-recognition-in-videos-on-something-something-v2StructVit-B-4-1–Top-1 Accuracy: 71.5
action-recognition-on-diving-48StructVit-B-4-1–Accuracy: 88.3