Accelerator-aware Neural Network Design using AutoML

TL;DR

This paper introduces hardware-aware neural architecture search tailored for the Edge TPU, resulting in models that deliver real-time image classification with high accuracy on low-power edge devices.

Contribution

It presents a novel AutoML-based approach for designing neural networks optimized specifically for Edge TPU hardware, enhancing performance and accuracy.

Findings

Models achieve real-time classification on Edge TPU.

Improved accuracy-latency tradeoff over existing mobile models.

Enhanced performance on Pixel 4's Edge TPU.

Abstract

While neural network hardware accelerators provide a substantial amount of raw compute throughput, the models deployed on them must be co-designed for the underlying hardware architecture to obtain the optimal system performance. We present a class of computer vision models designed using hardware-aware neural architecture search and customized to run on the Edge TPU, Google's neural network hardware accelerator for low-power, edge devices. For the Edge TPU in Coral devices, these models enable real-time image classification performance while achieving accuracy typically seen only with larger, compute-heavy models running in data centers. On Pixel 4's Edge TPU, these models improve the accuracy-latency tradeoff over existing SoTA mobile models.