Fast and Scalable 2D Convolutions and Cross-correlations for Processing Image Databases and Videos on CPUs
Fast and Scalable 2D Convolutions and Cross-correlations for Processing Image Databases and Videos on CPUs
复制标题
DOI：
10.1109/ssiai49293.2020.9094602
复制
发表时间：
2020-03
期刊：
影响因子：
--
通讯作者：
Cesar Carranza;D. Llamocca;M. Pattichis
中科院分区：
文献类型：
--
作者：
Cesar Carranza;D. Llamocca;M. Pattichis
卷积神经网络（cnn）在一些图像和视频分析任务中的主要使用需要仔细重新评估底层软件库，以便为大规模图像和视频数据库计算它们。我们专注于开发可以应用于大型图像数据库或大图像尺寸视频的方法。我们开发了一种通过使用基于矢量的内存I/O和优化的2D FFT库来最大化吞吐量的方法，这些库可以在所有可用的物理内核上运行。我们还展示了如何将任意大的图像分解成更小的、最优的块，这些块可以通过使用重叠和添加来有效地处理。我们的方法在5 × 5内核上优于Tensorflow，在11 × 11内核上优于Tensorflow。
The dominant use of Convolutional Neural Networks (CNNs) in several image and video analysis tasks necessitates a careful re-evaluation of the underlying software libraries for computing them for large-scale image and video databases. We focus our attention on developing methods that can be applied to large image databases or videos of large image sizes.We develop a method that maximizes throughput through the use of vector-based memory I/O and optimized 2D FFT libraries that run on all available physical cores. We also show how to decompose arbitrarily large images into smaller, optimal blocks that can be effectively processed through the use of overlap-and- add. Our approach outperforms Tensorflow for 5 × 5 kernels and significantly outperforms Tensorflow for 11 × 11 kernels.