DejaVid: Encoder-Agnostic Learned Temporal Matching for Video Classification

Darryl Ho, Samuel Madden

In recent years, large transformer-based video encoder models have greatly advanced state-of-the-art performance on video classification tasks. However, these large models typicall…