A Survey of Methods for Low-Power Deep Learning and Computer Vision
It addresses the problem of deploying large DNNs on resource-constrained devices for applications in mobile and embedded systems, but it is a survey paper, so it is incremental in summarizing existing research.
This paper surveys methods for reducing the energy, computation, and memory requirements of deep neural networks (DNNs) in computer vision, enabling deployment on low-power devices, and categorizes techniques like quantization, pruning, and architecture search while analyzing their trade-offs.
Deep neural networks (DNNs) are successful in many computer vision tasks. However, the most accurate DNNs require millions of parameters and operations, making them energy, computation and memory intensive. This impedes the deployment of large DNNs in low-power devices with limited compute resources. Recent research improves DNN models by reducing the memory requirement, energy consumption, and number of operations without significantly decreasing the accuracy. This paper surveys the progress of low-power deep learning and computer vision, specifically in regards to inference, and discusses the methods for compacting and accelerating DNN models. The techniques can be divided into four major categories: (1) parameter quantization and pruning, (2) compressed convolutional filters and matrix factorization, (3) network architecture search, and (4) knowledge distillation. We analyze the accuracy, advantages, disadvantages, and potential solutions to the problems with the techniques in each category. We also discuss new evaluation metrics as a guideline for future research.