Algorithmic Performance-Accuracy Trade-off in 3D Vision Applications Using HyperMapper
Luigi Nardi1, Bruno Bodin2, Sajad Saeedi1, Emanuele Vespa1, Andrew J. Davison1, Paul H. J. Kelly1 1Department of Computing, Imperial College London London, UK {l.nardi, s.saeedi, e.vespa14, ajd, 2Institute for Computing Systems Architecture, The University of Edinburgh Edinburgh, Scotland
Abstract
In this paper we investigate an emerging application, 3D scene understanding, likely to be significant in the mobile space in the near future. The goal of this exploration is to reduce execution time while meeting our quality of result objectives. In previous work, we showed for the first time that it is possible to map this application to power constrained embedded systems, highlighting that decision choices made at the algorithmic design-level have the most significant impact. As the algorithmic design space is too large to be exhaustively evaluated, we use a previously introduced multi-objective random forest active learning prediction framework dubbed HyperMapper, to find good algorithmic designs. We show that HyperMapper generalizes on a recent cutting edge 3D scene understanding algorithm and on a modern GPU-based computer architecture. HyperMapper is able to beat an expert human hand-tuning the algorithmic parameters of the class of computer vision applications taken under consideration in this paper automatically. In addition, we use crowd-sourcing using a 3D scene understanding Android app to show that the Pareto front obtained on an embedded system can be used to accelera
中文速览
移动设备上的实时三维场景理解（3D scene understanding）计算量极大，如何在功耗受限的嵌入式平台上又快又准地运行，是一个核心难题。研究者使用名为 HyperMapper 的多目标主动学习框架（multi-objective active learning framework），对 KinectFusion 和 ElasticFusion 两种主流稠密 SLAM 算法的算法级参数进行自动化设计空间探索，找出在运行速度与定位精度之间权衡最优的帕累托前沿（Pareto front）配置集合。实验结果显示，HyperMapper 不仅超越了领域专家的手动调参效果——在 ElasticFusion 上实现了最高 1.52 倍加速并将定位误差降低一半——还通过众包实验证明，在嵌入式设备上学到的最优配置可直接迁移至 83 款安卓手机和平板，带来 2 倍至逾 12 倍的加速。这项工作的意义在于，它提供了一套可跨硬件平台泛化、无需人工干预的自动调优方法，为将高算力计算机视觉应用落地到日常移动设备铺平了道路。
原文 arXiv:1702.00505;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1702.00505v2