High-efficiency and high-speed germanium photodetector enabled by multire...
Growth of high quality InSb thin films on GaAs substrates by molecular be...
A Low-Cost High-Speed Object Tracking VLSI System Based on Unified Textur...
The influence of temperature and TMGa flow rate on the quality of p-GaN
Modify Cd3As2 nanowires with sulfur to fabricate self-powered NIR photode...
Deformation Reduction in Laser Cladding of Sheet Metals by Heat Recycling
A review of research on co-training
Phonon renormalization in reconstructed MoS2 moire superlattices
Axiotaxy driven growth of belt-shaped InAs nanowires in molecular beam ep...
Improving the homogeneity and quality of InGaN/GaN quantum well exhibitin...

A 460 GOPS/W Improved Mnemonic Descent Method-Based Hardwired Accelerator for Face Alignment



Author(s): Mo, HY (Mo, Huiyu); Liu, LB (Liu, Leibo); Zhu, WP (Zhu, Wenping); Li, Q (Li, Qiang); Yin, SY (Yin, Shouyi); Wei, SJ (Wei, Shaojun)

Source: IEEE TRANSACTIONS ON MULTIMEDIA Volume: 23 Pages: 1122-1135 DOI: 10.1109/TMM.2020.2993943 Published: 2021

Abstract: The mnemonic descent method (MDM) algorithm is the first end-to-end recurrent convolutional system for high-accuracy face alignment. However, the heavy computational complexity and high memory access demands make it difficult to satisfy the requirements of real-time applications. To address this problem, an improved MDM (I-MDM) algorithm is proposed for efficient hardware implementation based on several hardware-oriented optimizations. First, a patch merging mechanism is introduced to dynamically cluster and eliminate redundant landmarks, which significantly reduces computational complexity with minimal accuracy loss. Second, a dedicated convolutional layer is inserted to halve the number of computations and memory access of the subsequent fully connected layer, yielding a 4.42% decrease in the failure rate. Third, a lightweight preprocessing method named dual regressors is proposed to reinitialize face images, which can greatly improve the overall accuracy. Moreover, compared with a similar method, the DR method can reduce computations and memory storage by nearly 99.9%. Overall and compared with the MDM algorithm, I-MDM not only reduces the number of computations by 23.5% but also decreases the failure rate by 17.9% on the 300 W test set. Based on the proposed I-MDM algorithm, an I-MDM-based hardwired accelerator is presented using the TSMC 65 nm CMOS process. First, compared with similar solutions, the gradient calculation operation is rearranged and loaded pixels are reused in the HoG feature extraction to eliminate all division operations and 25% off-chip memory access. Second, patch-independent central activations are used to enable patch-level pipelined operations, yielding a 2x acceleration in the overall process. This accelerator achieves 460 GOPS/W energy efficiency at 330 MHz, which is 38x higher than the most recent face alignment accelerator with the same process.

Accession Number: WOS:000623420300023

Author Identifiers:

Author        Web of Science ResearcherID        ORCID Number

Mo, Huiyu                  0000-0002-3373-7178

ISSN: 1520-9210

eISSN: 1941-0077

Full Text: https://ieeexplore.ieee.org/document/9091213


北京市海淀区清华东路甲35号 北京912信箱 (100083)





版权所有 中国科学院半导体研究所

备案号:,京ICP备05085259-1号 京公网安备110402500052 中国科学院半导体所声明