Fast deep neural networks for image processing using posits and ARM scalable vector extension
Fast deep neural networks for image processing using posits and ARM scalable vector extension
复制标题
DOI：
10.1007/s11554-020-00984-x
复制
发表时间：
2020-05
影响因子：
3
通讯作者：
M. Cococcioni;Federico Rossi;E. Ruffaldi;S. Saponara
中科院分区：
文献类型：
--
作者：
M. Cococcioni;Federico Rossi;E. Ruffaldi;S. Saponara
随着实时约束下汽车图像处理和计算机视觉的出现，对快速且架构优化的算术运算的需求至关重要。人们开始探索真实的数的替代和有效表示，其中，最近引入的正数系统是非常有前途的。此外，随着特定于架构的数学库的实现完全针对单指令多数据（SIMD）引擎，为深度神经网络框架提供的加速正在增加。在本文中，我们提出了一些核心的图像处理操作的实现，利用嵌入式算法和ARM的可扩展向量扩展SIMD引擎。此外，我们还介绍了实时图像处理在自动驾驶场景中的应用，并在tinyDNN深度神经网络（DNN）框架上展示了基准测试。
With the advent of image processing and computer vision for automotive under real-time constraints, the need for fast and architecture-optimized arithmetic operations is crucial. Alternative and efficient representations for real numbers are starting to be explored, and among them, the recently introduced positnumber system is highly promising. Furthermore, with the implementation of the architecture-specific mathematical library thoroughly targeting single-instruction multiple-data (SIMD) engines, the acceleration provided to deep neural networks framework is increasing. In this paper, we present the implementation of some core image processing operations exploiting the posit arithmetic and the ARM scalable vector extension SIMD engine. Moreover, we present applications of real-time image processing to the autonomous driving scenario, presenting benchmarks on the tinyDNN deep neural network (DNN) framework.