Resolving Temporal Misalignment in the Gait Recognition domain
Sprache des Vortragstitels:
Englisch
Original Tagungtitel:
OAGM Workshop 2023 | Patterns in One Health
Sprache des Tagungstitel:
Englisch
Original Kurzfassung:
We propose a novel end-to-end approach of combining model-based and appearance-based methods to circumvent the temporal misalignment problem in the gait recognition domain. Specifically, we locate and incorporate appearance-based features into a graph-based model. Features are extracted from the GREW dataset, which consists of real world gait data. Our approach overcomes the temporal misalignment problem, a distinction from existing works that alleviate but not completely circumvent the issue.
Forschungs-