Detection of robbery-related concepts using deep learning
Detecting robbery-related concepts or any particular violent scenes in videos is one of the most fundamental on-going work in the world of computer vision. While it is evident that there are more discovery and improvements of such detection task especially in the realm of fully supervised settings,...
| Main Author: | |
|---|---|
| Format: | Final Year Project / Dissertation / Thesis |
| Published: |
2020
|
| Subjects: | |
| Online Access: | http://eprints.utar.edu.my/3940/ http://eprints.utar.edu.my/3940/1/17ACB00362_FYP.pdf |
| _version_ | 1848886031281029120 |
|---|---|
| author | Vivaaindrean, Ng Shamir Ng |
| author_facet | Vivaaindrean, Ng Shamir Ng |
| author_sort | Vivaaindrean, Ng Shamir Ng |
| building | UTAR Institutional Repository |
| collection | Online Access |
| description | Detecting robbery-related concepts or any particular violent scenes in videos is one of the most fundamental on-going work in the world of computer vision. While it is evident that there are more discovery and improvements of such detection task especially in the realm of fully supervised settings, the acquisition of labelled training data at video’s temporal-level is not sensible. We instead tackle this problem by proposing two novel approaches – MIL-Ranking as well as TAL. At its very core, both aforementioned methods only necessitates ground-truth at video-level, instead of temporallevel. We show that the implementation of MIL and TAL approaches on the huge-scale UCF-Crime dataset demonstrates their capabilities in detecting violent-related concepts at video’s temporal-level. |
| first_indexed | 2025-11-15T19:32:01Z |
| format | Final Year Project / Dissertation / Thesis |
| id | utar-3940 |
| institution | Universiti Tunku Abdul Rahman |
| institution_category | Local University |
| last_indexed | 2025-11-15T19:32:01Z |
| publishDate | 2020 |
| recordtype | eprints |
| repository_type | Digital Repository |
| spelling | utar-39402021-01-07T08:02:29Z Detection of robbery-related concepts using deep learning Vivaaindrean, Ng Shamir Ng Q Science (General) Detecting robbery-related concepts or any particular violent scenes in videos is one of the most fundamental on-going work in the world of computer vision. While it is evident that there are more discovery and improvements of such detection task especially in the realm of fully supervised settings, the acquisition of labelled training data at video’s temporal-level is not sensible. We instead tackle this problem by proposing two novel approaches – MIL-Ranking as well as TAL. At its very core, both aforementioned methods only necessitates ground-truth at video-level, instead of temporallevel. We show that the implementation of MIL and TAL approaches on the huge-scale UCF-Crime dataset demonstrates their capabilities in detecting violent-related concepts at video’s temporal-level. 2020-05-15 Final Year Project / Dissertation / Thesis NonPeerReviewed application/pdf http://eprints.utar.edu.my/3940/1/17ACB00362_FYP.pdf Vivaaindrean, Ng Shamir Ng (2020) Detection of robbery-related concepts using deep learning. Final Year Project, UTAR. http://eprints.utar.edu.my/3940/ |
| spellingShingle | Q Science (General) Vivaaindrean, Ng Shamir Ng Detection of robbery-related concepts using deep learning |
| title | Detection of robbery-related concepts using deep learning |
| title_full | Detection of robbery-related concepts using deep learning |
| title_fullStr | Detection of robbery-related concepts using deep learning |
| title_full_unstemmed | Detection of robbery-related concepts using deep learning |
| title_short | Detection of robbery-related concepts using deep learning |
| title_sort | detection of robbery-related concepts using deep learning |
| topic | Q Science (General) |
| url | http://eprints.utar.edu.my/3940/ http://eprints.utar.edu.my/3940/1/17ACB00362_FYP.pdf |