Investigation of Neural Network Detecting of Construction Waste by Photos
Вантажиться...
Дата
2025
Назва журналу
Номер ISSN
Назва тому
Видавець
Анотація
This paper presents a practice-oriented investigation of neural architectures for visual recognition of construction and demolition (C&D) waste in realistic conditions. We study a two-stage pipeline that first localizes candidate fragments on a scene and then assigns materialspecific labels, emphasizing data curation, restrained augmentation, and evaluation protocols aligned with production constraints. Results indicate that coupling a single-shot detector with per class residual classifiers improves recall on visually confusable materials while preserving precision, offering a pragmatic path toward reliable sorting-line deployment.
Опис
Ключові слова
Бібліографічний опис
Mazurets O., Ovcharuk O., Vit R. Investigation of Neural Network Detecting of Construction Waste by Photos. Information Technology and Implementation (Satellite). Proceedings 12th International Conference. November 21, 2025. Kyiv, Ukraine. 2025. Pp. 96-97.