An Empirical Study of Person Re-Identification with Attributes

Abstract

Person re-identification aims to identify a person from an image collection, given one image of that person as the query. There is, however, a plethora of real-life scenarios where we may not have a priori library of query images and therefore must rely on information from other modalities. In this paper, an attribute-based approach is proposed where the person of interest (POI) is described by a set of visual attributes, which are used to perform the search. We compare multiple algorithms and analyze how the quality of attributes impacts the performance. While prior work mostly relies on high precision attributes annotated by experts, we conduct a human-subject study and reveal that certain visual attributes could not be consistently described by human observers, making them less reliable in real applications. A key conclusion is that the performance achieved by non-expert attributes, instead of expert-annotated ones, is a more faithful indicator of the status quo of attribute-based approaches for person re-identification.

Attribute Memory Transfer Network for Unsupervised Cross-Domain Person Re-Identification

“…Pedestrian attributes, such as wearable, haircut, shoes-style, accessories etc, have been introduced in the detailed feature learning in person re-ID researches [27], [28], [35]. Shree et al [28] describes the person of interest by a series of visual attributes in the person re-ID task, which is that the performance achieved by non-expert attributes to indicate the pedestrian status. Wang et al [35] proposes to incrementally generate deep hidden attributes for newly uncovered annotations in person re-ID, with an auto-encoder model to mine latent information in an unsupervised manner.…”

Interactive Natural Language-Based Person Search

“…To this end, we opt for five query questions about an image, each asking about certain appearance aspect of a person. Table II shows the questions, which are motivated from the recent work in attribute-based re-ID [22]. We randomly selected 400 images of 360 people from the test set of CUHK-PEDES dataset, and conducted a survey to label the images with answers to the query questions (named CUHK-QA).…”