Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
4947729 | Neurocomputing | 2017 | 32 Pages |
Abstract
Chinese face reading has demonstrated the often satisfying capabilities to tell the characteristics (mostly exaggerated as fortune) of a person by reading his/her face, i.e. understanding the fine-grained facial attributes (e.g., length of nose, single/double-fold eyelid, density of eyebrows, etc.). Thus, a smart face reading system should estimate the fine-grained facial attributes well. Therefore, In this paper, we first study the fine-grained facial attribute estimation problem and propose a novel deep convolutional network equipped with a new facial region pooling layer (called FRP-net), to accurately estimate the fine-grained facial attributes. To capture the characteristics of fine-grained facial attributes, the embedded FRP layer implements the pooling operation on the searched facial region windows (locates the region of each facial attribute) instead of the commonly-used sliding windows. Further, we push the proposed fine-grained facial attribute estimation method into the face reading problem and present a computational face reader system to automatically infer the characteristics of a person based on his/her face. For example, it can estimate the attractive and easy-going characteristics of a Chinese person from his/her big eyes according to the Chinese anthroposcopy literature. The experimental results on facial attribute estimation demonstrate the superiority of the proposed FRP-net compared to the baselines, and the qualitative and quantitative evaluations on face reading validate the excellence of the presented face reader system.
Keywords
Related Topics
Physical Sciences and Engineering
Computer Science
Artificial Intelligence
Authors
Xiangbo Shu, Yunfei Cai, Liu Yang, Liyan Zhang, Jinhui Tang,