| Title |
Clip-level suspicious activity detection in retail surveillance videos using human–object interaction and temporal modelling |
| Translation of Title |
Įtartinos veiklos aptikimas mažmeninės prekybos stebėjimo vaizdo įrašuose, taikant žmogaus–objekto sąveikos ir laiko sekų modeliavimo metodus. |
| Authors |
Sohaib, Muhammad |
| Full Text |
|
| Pages |
69 |
| Keywords [eng] |
shoplifting detection ; suspicious activity detection ; human–object interaction |
| Abstract [eng] |
Retail theft (shoplifting) has been a significant cause of financial loss to retailers in various parts of the world and has inspired the development of smarter video surveillance systems that can automatically alert suspicious activity. The proposed thesis is based on the hypothesis of determining whether the use of human-object interaction (HOI) semantics in temporal anomaly detection can significantly enhance detection quality over person-detection statistical baselines. This study is centred on retail surveillance video clip-level anomaly detection and measures performance based on standardised ranking measures. |
| Dissertation Institution |
Kauno technologijos universitetas. |
| Type |
Master thesis |
| Language |
English |
| Publication date |
2026 |
