Submission Deadline: 31 January 2025

========================

=== Call for papers ===

Pedestrian attributes recognition and person re-identification from
images and videos is nowadays a relevant problem in several real
applications, such as forensics, digital signage, social robotics,
business intelligence, people tracking and multi-camera person
re-identification.
In terms of neural network architecture, there is still a limited use of
visual attention mechanisms, which could allow for more accurate and
efficient part localization and recognition; furthermore, recent
advanced fully convolutional network architectures or based on
transformers could be explored. Newly designed loss functions could be
adopted to deal with unbalanced data, while multi-task learning
approaches may represent an excellent solution to exploit the
interdependencies between the attributes while maintaining the
processing time unchanged as the number of attributes increases.
The efficiency is another aspect that is important to consider, since
fast person re-identification is crucial for people tracking and
multi-camera person re-identification in crowded scenarios such stations
or airports; to this aim, in addition to multi-task architectures,
end-to-end solutions for people detection, re-identification and
tracking would be of great interest to the scientific community.
The robustness of these methods should be investigated in various
background, resolution, and illumination conditions, considering that
these variations can interfere with the recognition of pedestrian
attributes and affect the performance of person re-identification;
datasets and benchmark soliciting the algorithms in the wild would be a
significant contribution to this aim. To achieve this robustness,
multi-frame or multi-modal inputs may be considered; video-based or
multi-sensor (RGB camera, thermal camera, depth camera, LIDAR)
pedestrian attribute recognition and person re-identification approaches
could exploit the additional information to improve accuracy and
robustness of real applications. Finally, it is worth mentioning that
the very recent Pedestrian Attribute Recognition contest (PAR 2023) was
won by a method based on Visual Question Answering; this surprising and
impressive result suggests that the investigation of this type of
methods or other foundation models can be a very promising line of
research in this field.
Considering the relevance of the topic for the possible exploitation in
real applications and all the above-mentioned possible directions of
research and improvements of existing algorithms, the special issue has
the goal of collecting innovative scientific papers to advance the state
of the art in the recognition of pedestrian attributes and person
re-identification. The topic and the innovative contributions in this
field are aligned with the aims and scope of the journal and, thus, can
be relevant for its audience and readership.

Topics of interest of the proposed Special Issue, related to pedestrian
attribute recognition and person re-identification, are but not limited
to:

• Machine learning algorithms for human attribute recognition
• Deep learning models for people detection and classification
• Appearance based people tracking
• Multi-camera people re-identification
• Multi-frame person retrieval
• Multi-modal people detection and re-identification
• Visual question answering applied to human attributes
• New datasets and benchmark for pedestrian attribute recognition and
person re-identification
• Novel applications and case studies in surveillance scenarios

========================

=== Guest editors ===

Modesto Castrillon-Santana (University of Las Palmas De Gran Canaria,
Spain)
Antonio Greco (University of Salerno, Italy)
Nicolai Petkov (University of Groningen, The Netherlands)
Bruno Vento (University of Napoli Federico II, Italy)