On the fly Deep Neural Network Optimization Control for Low-Power Computer Vision

TL;DR

This paper introduces AdaptiveActivation, a run-time technique for DNNs that dynamically balances accuracy and energy consumption on edge devices without re-training, enhancing deployability across diverse hardware.

Contribution

The paper proposes a novel, re-training-free method to adapt DNN accuracy and efficiency during run-time using a hyper-parameter controlling activation output range.

Findings

Achieves within 1.5% accuracy of baseline on edge devices.

Reduces memory usage by 10-38% compared to existing techniques.

Enables flexible accuracy-efficiency trade-offs in diverse edge environments.

Abstract

Processing visual data on mobile devices has many applications, e.g., emergency response and tracking. State-of-the-art computer vision techniques rely on large Deep Neural Networks (DNNs) that are usually too power-hungry to be deployed on resource-constrained edge devices. Many techniques improve the efficiency of DNNs by using sparsity or quantization. However, the accuracy and efficiency of these techniques cannot be adapted for diverse edge applications with different hardware constraints and accuracy requirements. This paper presents a novel technique to allow DNNs to adapt their accuracy and energy consumption during run-time, without the need for any re-training. Our technique called AdaptiveActivation introduces a hyper-parameter that controls the output range of the DNNs' activation function to dynamically adjust the sparsity and precision in the DNN. AdaptiveActivation can be applied to any existing pre-trained DNN to improve their deployability in diverse edge environments. We conduct experiments on popular edge devices and show that the accuracy is within 1.5% of the baseline. We also show that our approach requires 10%--38% less memory than the baseline techniques leading to more accuracy-efficiency tradeoff options