Real-Time Video Inference on Edge Devices via Adaptive Model Streaming

TL;DR

This paper introduces Adaptive Model Streaming (AMS), a method that enables real-time video inference on edge devices by dynamically adapting lightweight models through remote training and online knowledge distillation, achieving high accuracy and low latency.

Contribution

The paper proposes AMS, a novel approach for over-the-network model adaptation that improves lightweight model performance for video inference on edge devices.

Findings

Achieves 0.4-17.8% mIoU improvement over pre-trained models.

Runs at 30 fps with 40 ms latency on a Samsung Galaxy S10+.

Uses less than 300 Kbps bandwidth for model updates.

Abstract

Real-time video inference on edge devices like mobile phones and drones is challenging due to the high computation cost of Deep Neural Networks. We present Adaptive Model Streaming (AMS), a new approach to improving performance of efficient lightweight models for video inference on edge devices. AMS uses a remote server to continually train and adapt a small model running on the edge device, boosting its performance on the live video using online knowledge distillation from a large, state-of-the-art model. We discuss the challenges of over-the-network model adaptation for video inference, and present several techniques to reduce communication cost of this approach: avoiding excessive overfitting, updating a small fraction of important model parameters, and adaptive sampling of training frames at edge devices. On the task of video semantic segmentation, our experimental results show 0.4--17.8 percent mean Intersection-over-Union improvement compared to a pre-trained model across several video datasets. Our prototype can perform video segmentation at 30 frames-per-second with 40 milliseconds camera-to-label latency on a Samsung Galaxy S10+ mobile phone, using less than 300 Kbps uplink and downlink bandwidth on the device.