Attention in Multimodal Neural Networks for Person Re-identification

In spite of increasing interest from the research commu-
nity, person re-identification remains an unsolved problem.
Correctly deciding on a true match by comparing images
of a person, captured by several cameras, requires extrac-
tion of discriminative features to counter challenges such as
changes in lighting, viewpoint and occlusion. Besides de-
vising novel feature descriptors, the setup can be changed
to capture persons from an overhead viewpoint rather than
a horizontal. Furthermore, additional modalities can be
considered that are not affected by similar environmental
changes as RGB images. In this work, we present a Multi-
modal ATtention network (MAT) based on RGB and depth
modalities. We combine a Convolution Neural Network with
an attention module to extract local and discriminative fea-
tures that are fused with globally extracted features. At-
tention is based on correlation between the two modalities
and we finally also fuse RGB and depth features to generate
a joint multilevel RGB-D feature. Experiments conducted
on three datasets captured from an overhead view show the
importance of attention, increasing accuracies by 3.43%,
2.01% and 2.13% on OPR, DPI-T and TVPR, respectively.