FlexDNN: Input-Adaptive On-Device Deep Learning for Efficient Mobile Vision
FlexDNN: Input-Adaptive On-Device Deep Learning for Efficient Mobile Vision
复制标题
DOI：
10.1109/sec50012.2020.00014
复制
发表时间：
2020-11
期刊：
影响因子：
--
通讯作者：
Biyi Fang;Xiao Zeng;Faen Zhang;Hui Xu;Mi Zhang
中科院分区：
文献类型：
--
作者：
Biyi Fang;Xiao Zeng;Faen Zhang;Hui Xu;Mi Zhang
由深度神经网络（dnn）的最新进展驱动的移动视觉系统正在实现广泛的设备上视频分析应用。考虑到移动系统受限于有限的资源，减少dnn的资源需求对于实现这些应用的全部潜力至关重要。在本文中，我们提出了FlexDNN，这是一种基于输入自适应dnn的框架，用于高效的设备上视频分析。为了实现这一点，FlexDNN考虑了移动视频的内在动态性，并根据输入视频帧的难易程度动态调整其模型复杂度，以达到计算效率。FlexDNN解决了现有系统的主要缺点，并推动了最先进的技术向前发展。我们使用FlexDNN构建了三个具有代表性的设备上视频分析应用程序，并评估了其在移动CPU和GPU平台上的性能。我们的结果表明，FlexDNN在精度、每帧平均CPU/GPU处理时间、帧丢帧率和能耗方面显著优于目前的方法。
Mobile vision systems powered by the recent advancement in Deep Neural Networks (DNNs) are enabling a wide range of on-device video analytics applications. Considering mobile systems are constrained with limited resources, reducing resource demands of DNNs is crucial to realizing the full potential of these applications. In this paper, we present FlexDNN, an input-adaptive DNN-based framework for efficient on-device video analytics. To achieve this, FlexDNN takes the intrinsic dynamics of mobile videos into consideration, and dynamically adapts its model complexity to the difficulty levels of input video frames to achieve computation efficiency. FlexDNN addresses the key drawbacks of existing systems and pushes the state-of-the-art forward. We use FlexDNN to build three representative on-device video analytics applications, and evaluate its performance on both mobile CPU and GPU platforms. Our results show that FlexDNN significantly outperforms status quo approaches in accuracy, average CPU/GPU processing time per frame, frame drop rate, and energy consumption.