Clothing cosegmentation for recognizing people

Abstract

Reseachers have verified that clothing provides information about the identity of the individual. To extract features from the clothing, the clothing region first must be localized or segmented in the image. At the same time, given multiple images of the same person wearing the same clothing, we expect to improve the effectiveness of clothing segmentation. Therefore, the identity recognition and clothing segmentation problems are inter-twined; a good solution for one aides in the solution for the other.We build on this idea by analyzing the mutual information between pixel locations near the face and the identity of the person to learn a global clothing mask. We segment the clothing region in each image using graph cuts based on a clothing model learned from one or multiple images believed to be the same person wearing the same clothing. We use facial features and clothing features to recognize individuals in other images. The results show that clothing segmentation provides a significant improvement in recognition accuracy for large image collections, and useful clothing masks are simultaneously produced.A further significant contribution is that we introduce a publicly available consumer image collection where each individual is identified. We hope this dataset allows the vision community to more easily compare results for tasks related to recognizing people in consumer image collections.

Biometric Authentication

“…High correlation is symbolized by orange, and low by blue/green. In the categorical matrix, traits relating to head coverage (2) and (4) are highly correlated, as are the traits (15) and (18) relating to the description of items attached to the body. Clothing categories are well correlated for upper (5) and lower (9) body, as expected.…”

“…The majority of existing research employs computer vision algorithms and machine learning techniques to extract and use visual clothing descriptions in applications including: online person recognition [5,11]; semantic attributes for re-identification [13]; detecting and analysing semantic descriptions (labels) of clothing colours and types to supplement other bodily and facial soft attributes in automatic search and retrieval [9]; and utilizing some clothing attributes like colour [14] and style to improve the observation and retrieval at a distance in surveillance environments [2]. Even with images captured on different days, there remains sufficient information to compare and establish identity, since clothes are often re-worn or a particular individual may prefers a specific clothing style or colour [15]. Clothing descriptions like indicative colours and decorations could be utilized to supplement other behavioural biometrics like human motion pattern, hence they can form a biometric fingerprint that serves as a person's identifier [11].…”

Automatic Generation of Subject-Based Image Transitions

“…For this reason we used personal collections both for training and testing. The training set is a collection of faces detected in a private personal collection, while the whole system has been tested on a publicly available dataset [11] enabling future comparison.…”

“…We created a benchmark of videos based on the images in the Gallagher Collection Person Dataset [11]. The videos were realized using 24 frames per second and 5 image transitions with a running time of 1 second each.…”

An Automated System for Garment Texture Design Class Identification

“…Recently, garment related research has become popular [1][2][3][4][5]. Most of the works focus on the segmentation of garments from real life images.…”