Abstract:
One biometrics modality is gait. Most gait recognition methods use silhouettes or articulated body models. Confounding variables like clothing, carrying, and viewing angle degrade recognition performance in these methods. We propose GaitNet, a new AutoEncoder framework, to explicitly separate appearance, canonical, and pose features from RGB imagery. The LSTM integrates pose features over time as a dynamic gait feature and averages canonical features as a static one. Both are classifiers. We also collect a Frontal-View Gait (FVG) dataset to focus on frontal-view walking gait recognition, which is difficult due to its limited gait cues. Walking speed, carrying, and clothing are also important FVG variations. Our method outperforms the SOTA quantitatively, disentangles features qualitatively, and is computationally efficient. To demonstrate the advantages of gait biometrics identification in certain scenarios, such as long-distance/lower resolutions, cross viewing angles, we compare our GaitNet with state-of-the-art face recognition.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
