發(fā)布時間：2018-07-05 05:13

  本文選題：深度學習 + 卷積神經(jīng)網(wǎng)絡　； 參考：《廈門大學學報(自然科學版)》2017年01期

【摘要】：近年來,基于深度學習的人臉分析取得了巨大的進步,成為計算機視覺領(lǐng)域最為活躍的研究方向之一.為了進一步推動深度學習和人臉分析的研究,結(jié)合近年已發(fā)表的相關(guān)文獻,對基于深度學習的人臉分析技術(shù)進行綜述.首先,簡要概述深度學習及其發(fā)展歷史,并分析深度學習有效性原因.然后,按照任務目的的不同,將人臉分析分成了人臉檢測、人臉關(guān)鍵點檢測、人臉識別、人臉屬性識別等任務進行詳細的介紹和討論,重點分析各種任務現(xiàn)階段存在的主要問題.接著,介紹人臉分析中常用的人臉數(shù)據(jù)庫.最后,討論深度學習和人臉分析面臨的主要挑戰(zhàn),并給出結(jié)論.
[Abstract]:In recent years, face analysis based on deep learning has made great progress and become one of the most active research directions in the field of computer vision. In order to further promote the research of depth learning and face analysis, combined with the related literature published in recent years, this paper summarizes the technology of face analysis based on depth learning. First of all, it briefly summarizes the history of deep learning and its development, and analyzes the reasons for the effectiveness of depth learning. Then, according to the different purpose of the task, face analysis is divided into face detection, face key point detection, face recognition, face attribute recognition and other tasks for detailed introduction and discussion. Focus on the analysis of various tasks at the present stage of the main problems. Then, the face database commonly used in face analysis is introduced. Finally, the main challenges of depth learning and face analysis are discussed and the conclusions are given.
【作者單位】： 廈門大學信息科學與技術(shù)學院福建省智慧城市感知與計算重點實驗室;
【基金】：國家自然科學基金(61571379,61472334)
【分類號】：TP391.41
，

本文編號：2099061