Bayesian Gait-Based Gender Identification (BGGI) Network on Individuals Wearing Loosely Fitted Clothing

Amarjot Singh, Aman Kumar, Anisha Jain; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2019, pp. 0-0

Abstract


Suspicious individuals often attempt to hide their identity to avoid detection by safety and security systems. Wearing clothes of the opposite gender is one of the several techniques used by these individuals. Several promising attempts have been made to recognize gender using gait recognition. However, these systems only focused on recognizing gender for individuals who wore tightly fitting attire which made it easier to detect the body joints further making it possible to differentiate both genders. In this work, we attempt to solve a challenging real-world problem faced by security agencies in which the individuals mask their identity by wearing loosely fitted clothes (LFC) of the opposite gender. LFC makes it difficult to locate the body joints in effect making the gender classification, in this situation, a complicated problem. We propose a Bayesian Gait-based Gender Identification (BGGI) technique that is used for gender recognition in LFC conditions, in dense real-world videos. This research releases the loosely fitted clothes individuals (LFCI) dataset used for training the deep network. This may encourage researchers interested in using deep learning for this task. The pose estimation and gender recognition achieve great performance with state-of-the-art techniques.

Related Material


[pdf]
[bibtex]
@InProceedings{Singh_2019_ICCV,
author = {Singh, Amarjot and Kumar, Aman and Jain, Anisha},
title = {Bayesian Gait-Based Gender Identification (BGGI) Network on Individuals Wearing Loosely Fitted Clothing},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops},
month = {Oct},
year = {2019}
}