大理大学学报 ›› 2021, Vol. 6 ›› Issue (12): 23-28.

• 物理学 • 上一篇    下一篇

基于Faster-RCNN的民族服饰识别系统的设计与实现

  

  1. 大理大学工程学院,云南大理 671003
  • 收稿日期:2021-07-05 出版日期:2021-12-15 发布日期:2022-01-14
  • 作者简介:赵恩铭,副教授,博士,主要从事目标检测、嵌入式研究。
  • 基金资助:
    国家自然科学基金项目(62065001;61761048);云南省地方本科高校(部分)基础研究联合专项资金项目(2019FH001-066)

Design and Implementation of Ethnic Costume Recognition System Based on Faster-RCNN

  1. College of Engineering, Dali University, Dali, Yunnan 671003, China
  • Received:2021-07-05 Online:2021-12-15 Published:2022-01-14

摘要: 民族服饰文化是中华民族的瑰宝,民族服饰的辨别对于弘扬民族文化起到了积极的促进作用。在目标识别领域,卷积神经网络具有突出的表现。为辨别传统民族服饰,设计并实现了基于Faster-RCNN算法的民族服饰识别系统。系统实现了白族、苗族和蒙古族服饰的辨识,能够输出目标的分类标签、置信度和位置坐标信息,同时该系统能够对多目标服饰识别。经过测试,识别模型的MAP值为67.88%,达到了更好的民族服饰识别效果。

关键词: Faster-RCNN, 民族服饰识别, 卷积神经网络

Abstract:

Ethnic costumes is a treasure of the Chinese nation and the identification of ethnic costumes has played a positive role in promoting ethnic cultures. In the field of target recognition convolutional neural networks have an outstanding performance. In order to distinguish traditional ethnic costumes a recognition system based on Faster-RCNN algorithm was designed and implemented. The system realizes the identification of Bai Miao and Mongolian costumes and can output target-classification labels confidence intervals and coordinate information. At the same time the system can recognize multi-target costumes. After testing the MAP value of the recognition model is 67.88% which achieves a better effect of ethnic costume recognition.

Key words: Faster-RCNN, ethnic costume recognition, convolutional neural networks

中图分类号: