Arena: A Patch-of-Interest ViT Inference Acceleration System for Edge-Assisted Video Analytics
This work addresses the computational and bandwidth limitations of visual foundation models for edge-assisted video analytics, offering a practical solution for real-time applications in adverse environments.
The paper tackles the challenge of applying Vision Transformers (ViT) to real-time video analytics on edge devices by proposing Arena, a system that accelerates inference through token pruning and adaptive keyframe switching, achieving up to 1.82× speedup and reducing bandwidth usage by up to 69% while maintaining high accuracy.
The advent of edge computing has made real-time intelligent video analytics feasible. Previous works, based on traditional model architecture (e.g., CNN, RNN, etc.), employ various strategies to filter out non-region-of-interest content to minimize bandwidth and computation consumption but show inferior performance in adverse environments. Recently, visual foundation models based on transformers have shown great performance in adverse environments due to their amazing generalization capability. However, they require a large amount of computation power, which limits their applications in real-time intelligent video analytics. In this paper, we find visual foundation models like Vision Transformer (ViT) also have a dedicated acceleration mechanism for video analytics. To this end, we introduce Arena, an end-to-end edge-assisted video inference acceleration system based on ViT. We leverage the capability of ViT that can be accelerated through token pruning by only offloading and feeding Patches-of-Interest to the downstream models. Additionally, we design an adaptive keyframe inference switching algorithm tailored to different videos, capable of adapting to the current video content to jointly optimize accuracy and bandwidth. Through extensive experiments, our findings reveal that Arena can boost inference speeds by up to 1.58\(\times\) and 1.82\(\times\) on average while consuming only 47\% and 31\% of the bandwidth, respectively, all with high inference accuracy.