幸运pk10

波兰科学院院士Tomasz Dietl教授谈磁性半导体
半导体学报2019年第8期——磁性半导体的机遇与挑战专_x005f_x005f_x005f_x005f_x005f_x005f_x005f_x005f_x005f_x005f_x005f_x005f_x005f_x005f_x005f_x005f_x005f_x005f_x005f_x005f_x0008_刊
A Fast and Universal RFID Tag Anti-Collision Algorithm for the Internet o...
Self-Powered Flexible Blood Oxygen Monitoring System Based on a Triboelec...
Performance Improvement of GaN Based Laser Diode Using Pd/Ni/Au Metalliza...
Enhancement of Heat Dissipation in Ultraviolet Light-Emitting Diodes by a...
Heterointerface-Driven Band Alignment Engineering and its Impact on Macro...
Highly Polarized Photoelectrical Response in vdW ZrS3 Nanoribbons
Strain compensated robust semiconductor saturable absorber mirror for fib...
官方微信
友情链接

A Face Alignment Accelerator Based on Optimized Coarse-to-Fine Shape Searching

2019-08-23

Authors: Liu, LB; Wang, Q; Zhu, WP; Mo, HY; Wang, TC; Yin, SY; Shi, YY; Wei, SJ

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY

Volume: 29 Issue: 8 Pages: 2467-2481 Published: AUG 2019 Language: English Document type: Article

DOI: 10.1109/TCSVT.2018.2867499

Abstract:

The coarse-to-fine shape searching (CFSS) framework is a recently developed algorithm that achieves relatively high accuracy in face alignment by alleviating the poor initialization problem facing traditional cascaded regression approaches. However, its high computational complexity and memory access demands make it difficult for CFSS to satisfy the requirements of real-time processing. To address this issue, a fast shape searching face alignment (F-SSFA) accelerator is presented based on the optimization of the CFSS algorithm and an efficient hardware implementation. First, the learning-based low-dimensional speeded-up robust features method, based on the correlations between the SURF features and the regression targets, is introduced to distill the feature set down to the only most distinct features to reduce the computing load. Second, the partial keypoints Euclidean distance and shape affine transformation are introduced to replace feature extraction and support vector machine classification, thereby accelerating the shape searching process. Compared with CFSS, F-SSFA achieves a 5.8x speedup while achieving similar accuracy. Moreover, a VLSI architecture is proposed to realize the fixed-point F-SSFA algorithm. Multiple descriptors located in adjacent regions are simultaneously generated in a single access to the corresponding image data. Therefore, repeated memory access operations are avoided. The optimal parameter configuration for hardware implementation is also exploited based on a tradeoff between accuracy and hardware performance. Simulated with TSMC 65-nm 1P8M technology within a 3.6 mm(2) area, a post-layout simulation shows that 700 fps can be achieved while consuming 300 mW at 200 MHz.

Full Text:



关于我们
下载视频观看
联系方式
通信地址

北京市海淀区清华东路甲35号 北京912信箱 (100083)

电话

010-82304210/010-82305052(传真)

E-mail

semi@semi.ac.cn

友情链接
版权所有 ? 幸运pk10

备案号:京ICP备05085259号 京公网安备110402500052