WildGait: Learning Gait Representations from Raw Surveillance Streams
The use of gait for person identification has important advantages such as being non-invasive, unobtrusive, not requiring cooperation and being less likely to be obscured compared to other biometrics. Existing methods for gait recognition require cooperative gait scenarios, in which a single person...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-12-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/21/24/8387 |
| _version_ | 1797500757370470400 |
|---|---|
| author | Adrian Cosma Ion Emilian Radoi |
| author_facet | Adrian Cosma Ion Emilian Radoi |
| author_sort | Adrian Cosma |
| collection | DOAJ |
| description | The use of gait for person identification has important advantages such as being non-invasive, unobtrusive, not requiring cooperation and being less likely to be obscured compared to other biometrics. Existing methods for gait recognition require cooperative gait scenarios, in which a single person is walking multiple times in a straight line in front of a camera. We address the challenges of real-world scenarios in which camera feeds capture multiple people, who in most cases pass in front of the camera only once. We address privacy concerns by using only motion information of walking individuals, with no identifiable appearance-based information. As such, we propose a self-supervised learning framework, WildGait, which consists of pre-training a Spatio-Temporal Graph Convolutional Network on a large number of automatically annotated skeleton sequences obtained from raw, real-world surveillance streams to learn useful gait signatures. We collected and compiled the largest pretraining dataset to date of anonymized walking skeletons called Uncooperative Wild Gait, containing over 38k tracklets of anonymized walking 2D skeletons. We make the dataset available to the research community. Our results surpass the current state-of-the-art pose-based gait recognition solutions. Our proposed method is reliable in training gait recognition methods in unconstrained environments, especially in settings with scarce amounts of annotated data. |
| first_indexed | 2024-03-10T03:08:26Z |
| format | Article |
| id | doaj.art-29b2838916e24bc586347237f43e9e18 |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-10T03:08:26Z |
| publishDate | 2021-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-29b2838916e24bc586347237f43e9e182023-11-23T10:30:46ZengMDPI AGSensors1424-82202021-12-012124838710.3390/s21248387WildGait: Learning Gait Representations from Raw Surveillance StreamsAdrian Cosma0Ion Emilian Radoi1Faculty of Automatic Control and Computer Science, University Politehnica of Bucharest, 060042 Bucharest, RomaniaFaculty of Automatic Control and Computer Science, University Politehnica of Bucharest, 060042 Bucharest, RomaniaThe use of gait for person identification has important advantages such as being non-invasive, unobtrusive, not requiring cooperation and being less likely to be obscured compared to other biometrics. Existing methods for gait recognition require cooperative gait scenarios, in which a single person is walking multiple times in a straight line in front of a camera. We address the challenges of real-world scenarios in which camera feeds capture multiple people, who in most cases pass in front of the camera only once. We address privacy concerns by using only motion information of walking individuals, with no identifiable appearance-based information. As such, we propose a self-supervised learning framework, WildGait, which consists of pre-training a Spatio-Temporal Graph Convolutional Network on a large number of automatically annotated skeleton sequences obtained from raw, real-world surveillance streams to learn useful gait signatures. We collected and compiled the largest pretraining dataset to date of anonymized walking skeletons called Uncooperative Wild Gait, containing over 38k tracklets of anonymized walking 2D skeletons. We make the dataset available to the research community. Our results surpass the current state-of-the-art pose-based gait recognition solutions. Our proposed method is reliable in training gait recognition methods in unconstrained environments, especially in settings with scarce amounts of annotated data.https://www.mdpi.com/1424-8220/21/24/8387gait recognitionpose estimationgraph neural networksself-supervised learning |
| spellingShingle | Adrian Cosma Ion Emilian Radoi WildGait: Learning Gait Representations from Raw Surveillance Streams Sensors gait recognition pose estimation graph neural networks self-supervised learning |
| title | WildGait: Learning Gait Representations from Raw Surveillance Streams |
| title_full | WildGait: Learning Gait Representations from Raw Surveillance Streams |
| title_fullStr | WildGait: Learning Gait Representations from Raw Surveillance Streams |
| title_full_unstemmed | WildGait: Learning Gait Representations from Raw Surveillance Streams |
| title_short | WildGait: Learning Gait Representations from Raw Surveillance Streams |
| title_sort | wildgait learning gait representations from raw surveillance streams |
| topic | gait recognition pose estimation graph neural networks self-supervised learning |
| url | https://www.mdpi.com/1424-8220/21/24/8387 |
| work_keys_str_mv | AT adriancosma wildgaitlearninggaitrepresentationsfromrawsurveillancestreams AT ionemilianradoi wildgaitlearninggaitrepresentationsfromrawsurveillancestreams |