Real-time face attributes recognition via HPGC: horizontal pyramid global convolution
- PDF / 1,363,406 Bytes
- 12 Pages / 595.276 x 790.866 pts Page_size
- 82 Downloads / 180 Views
SPECIAL ISSUE PAPER
Real‑time face attributes recognition via HPGC: horizontal pyramid global convolution Shimeng Yang1 · Fudong Nian2,3 · Yan Wang2 · Teng Li2 Received: 3 April 2019 / Accepted: 23 November 2019 © Springer-Verlag GmbH Germany, part of Springer Nature 2019
Abstract Recognizing face attributes in the wild is a challenging problem. With the development of embedded devices, smart phone and deep learning, face attributes recognition based on deep learning has raised increasing attention. Accuracy and efficiency are the two key-points in any application which uses these face attributes as an aid system. In most of the previous papers, multi-independency classifiers are proposed; and most of them just focus on accurate rate while neglecting efficiency. This paper proposes a horizontal pyramid global convolution (HPGC) module as feature mapping operator to extract more local information; designs a light-weight attribute convolution neural network (LACNN) combining with HPGC; and utilizes sigmoid cross entropy loss function for improving the accuracy and efficiency of the face attributes recognition model. Replacing full connection or global average pooling with the proposed HPGC module, we balance the accuracy performance and computation cost. As a result, we not only get high accuracy but also reduce the computational cost. Extensive experiments results on two widely used face attribute datasets, LFW and CelebA, demonstrate that our LACNN-HPGC framework achieves significantly improved efficiency compared with state-of-the-art lightweight models for face attributes recognition. Keywords Face attribute recognition · HPGC · CNN · Accuracy and efficiency · Sigmoid cross entropy loss
1 Introduction Human facial attributes have been extensively studied because of its relevant applications in the field of computer vision and pattern recognition [1–3]. Some attributes including gender, race, age, skin color, etc. are used for face recognition and verification successfully by Kumar et al. [4, 5] and Hu et al. [6]. Compared with deep learning-based face verification, face attributes-based methods are more efficient for many different tasks (image and video retrieval, law enforcement, human computer interaction applications). People of interest, who are criminal suspect or lost people, are always described by their identifiable facial attributes * Teng Li [email protected] Fudong Nian [email protected] 1
School of Computer Science and Technology, Anhui University, Hefei, China
2
Anhui University, Hefei, China
3
Hefei University, Hefei, China
such as big nose, high cheekbones, bushy eyebrows and so on. In large-scale low-resolution surveillance video face retrieval task, facial attribute-based methods are proved to be an excellent way. The face attributes classification models deployed on embedded devices (surveillance cameras, smart phones, etc.) are expected to be not only accurate but also small and efficient. Improving the accuracy and efficiency of face attributes classifiers is a challengi
Data Loading...
