Search academic texts

Information × Registration Number 2123U006677, Article popup.category Препринт Title popup.author Dobrovolskyy Nazar popup.publication 01-01-2023 popup.source_user Український католицький університет popup.source https://hdl.handle.net/20.500.14570/4822 popup.publisher Description Among the devastation of the Russian-Ukrainian war, the conquered territories became a breeding ground for unexploded mines and ammunition. The indiscrimi- nate use of these deadly weapons has caused countless civilian casualties and ham- pered recovery efforts in the affected regions. The urgent need to clear these areas requires effective and reliable detection of hidden mines and ordnance, which is challenging for human sappers due to the complexity and variability of the terrain and associated hazards. In this thesis, we propose solving this problem by building an object detection model using the YOLOv7 algorithm. We pre-collected and processed a large image dataset, created annotations for mines and projectiles, and trained and evaluated a model on this dataset using standard evaluation metrics. Our results show that our model provides high accuracy and efficiency in detecting and locating mines and projectiles in various scenarios. This research contributes to object detection and has practical implications for humanitarian demining operations. The model we developed integrates into ground remote robots for autonomous scanning and mapping of minefields, which reduces the risk of human casualties and accelerates the process of clearing territories. This thesis highlights the need to utilize new technology to address the critical issues of post-conflict rebuilding and human security. We believe our effort will encourage future study and development in this subject, ultimately contributing to peace and stability in the afflicted countries.. . . popup.nrat_date 2025-05-09 Close