Vision Based Few-Shot Railway Intrusion Detection Via Dual Detector and Contrastive Learning

Mrs. B. Lekhya, M.Tech,
Department of Electronics andcommunication engineering,
Annamacharya Institute of Technologyand Sciences, Tirupati,
bejawadalekhya@gmail.com
Kudimi Sai Lokesh, B.Tech, Department ofElectronics and communication engineering,
Annamacharya Institute of Technology andSciences, Tirupati,
kudimisailokesh@gmail.com

Allam Prathima, B.Tech,
Department of Electronics and communicationengineering, Annamacharya Institute of
Technology and Sciences, Tirupati,allamprathima2003@gmail.com
Varikuntla Sarath Kumar, B.Tech,
Department of Electronics and communicationengineering, Annamacharya Institute of
Technology and Sciences, Tirupati,varikuntlasarathkumar@gmail.com

ABSTRACT:Railway transportation systems require reliablemonitoring mechanisms to ensure track safety and prevent accidents caused by unexpected obstacles. The presence of humans, animals, or objects on railway tracks can lead to dangerous situationsandoperationaldisruptions.Recent improvements in computer vision have allowed automated surveillance systems to identifyobjects as they happen in real time. A vision based system is developed to detect railway intrusions using a dual- detector architecture
combined with few- shot learning and contrastive learning techniques.