An Enhanced YOLOV5 Model for Gateways Recognition in Heritage Buildings

Abstract

In India, gateways have been an integral part of architecture and have served as entrances to many historical buildings. These gateways are known for their unique design, intricate carvings, and beautiful ornamentation. The gateways of heritage buildings in India are not only significant architectural features but also have historical, cultural, and religious significance. At present time detecting gateways in heritage buildings is a difficult task for tourism agencies. To address the gateway recognition through real-time captured images, a novel-based heritage gateway recognition system is proposed through an enhanced ET-YOLOV5 object detector. The ET-YOLOV5 model uses the Resnet-50 as a feature extraction and spatial pyramid pooling model. The ETYOLOV5 model has been trained, tested, and validated on preprocessed 3000 heritage buildings image datasets. During the comparison, the ET-YOLOV5 increases the 9% mAP rate as compared to YOLOV5 and YOLOV4 for gateways recognition in heritage buildings of India.