Grouped Adaptive Loss Weighting for Person Search

Yanling TIAN, Dichen Li, Yunan Liu, Jian Yang, Shanshan Zhang

Abstract:

Person search is an integrated task of multiple sub-tasks such as
foreground/background classification, bounding box regression
and person re-identification. Therefore, person search is a typical
multi-task learning problem, especially when solved in an end-to-
end manner. Recently, some works enhance person search features
by exploiting various auxiliary information, e.g. person joint key-
points, body part position, attributes, etc., which brings in more
tasks and further complexifies a person search model. The inconsis-
tent convergence rate of each task could potentially harm the model
optimization. A straightforward solution is to manually assign dif-
ferent weights to different tasks, compensating for the diverse con-
vergence rates. However, given the special case of person search,
i.e. with a large number of tasks, it is impractical to weight the
tasks manually. To this end, we propose a Grouped Adaptive Loss
Weighting (GALW) method which adjusts the weight of each task
automatically and dynamically. Specifically, we group tasks accord-
ing to their convergence rates. Tasks within the same group share
the same learnable weight, which is dynamically assigned by con-
sidering the loss uncertainty. Experimental results on two typical
benchmarks, CUHK-SYSU and PRW, demonstrate the effectiveness
of our method.