ENHANCING DENSE VIDEO ANALYSIS TRANSFORMERS

Summary from the TLDR AI newsletter

A new Temporal Dilated Video Transformer (TDViT) has been developed to improve the analysis of dense video tasks, such as frame-by-frame video object detection.