人工智能培训

搜索

人工智能论文:FIRE SSD:边缘设备上基于宽火模块的单次探测器(Fire SSD: Wide Fire Modules based Single Shot

[复制链接]
632956335 发表于 2018-6-15 09:06:55 | 显示全部楼层 |阅读模式
632956335 2018-6-15 09:06:55 1795 0 显示全部楼层
人工智能论文:FIRE SSD:边缘设备上基于宽火模块的单次探测器(Fire SSD: Wide Fire Modules based Single Shot Detector on Edge Device)随着边缘计算的出现,越来越需要运行基于卷积神经网络的小型计算和热预算的小型边缘计算设备的对象检测,以应用于视频监控。为了解决这个问题,提出了有效的对象检测框架,如YOLO和SSD。但是,使用VGG16作为后端网络的基于SSD的对象检测不足以实现边缘设备的实时速度。为了进一步提高检测速度,后端网络被更高效的网络(如SqueezeNet和MobileNet)所取代。虽然速度大大提高,但它具有较低的准确性。在本文中,我们提出了一种名为Fire SSD的高效SSD。 FireSSD在Pascal VOC 2007测试装置上达到70.7mAP。 Fire SSD在低功耗主流CPU上实现了30.6FPS的速度,比SSD300快6倍,并且尺寸缩小了约4倍。 Fire SSD也实现了22.2FPS的集成GPU。
With the emergence of edge computing, there is an increasing need for runningconvolutional neural network based object detection on small form factor edgecomputing devices with limited compute and thermal budget for applications suchas video surveillance.To address this problem, efficient object detectionframeworks such as YOLO and SSD were proposed.However, SSD based objectdetection that uses VGG16 as backend network is insufficient to achieve realtime speed on edge devices.To further improve the detection speed, the backendnetwork is replaced by more efficient networks such as SqueezeNet andMobileNet.Although the speed is greatly improved, it comes with a price oflower accuracy.In this paper, we propose an efficient SSD named Fire SSD.FireSSD achieves 70.7mAP on Pascal VOC 2007 test set.Fire SSD achieves the speedof 30.6FPS on low power mainstream CPU and is about 6 times faster than SSD300and has about 4 times smaller model size.Fire SSD also achieves 22.2FPS onintegrated GPU.人工智能论文:FIRE SSD:边缘设备上基于宽火模块的单次探测器(Fire SSD: Wide Fire Modules based Single Shot Detector on Edge Device) zJjOcjV2fZqUCUf6.jpg
URL地址:https://arxiv.org/abs/1806.05363     ----pdf下载地址:https://arxiv.org/pdf/1806.05363    ----人工智能论文:FIRE SSD:边缘设备上基于宽火模块的单次探测器(Fire SSD: Wide Fire Modules based Single Shot Detector on Edge Device)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

632956335当前离线
新手上路

查看:1795 | 回复:0

快速回复 返回顶部 返回列表