Analysis of Domestic Building Detection Based on the YOLO

Building detection is an important task for land monitoring, urban planning, and illegal building. Recently, research is increasing to apply deep learning for building detection. However, prior studies are difficult to apply domestically because the dataset has different geographic and building characteristics. In this paper, we created a dataset for domestic building detection by collecting high-resolution satellite imagery. And, the deep learning model (YOLOv5) that achieved SOTA (state-of-the-art) was trained with our dataset and detected building. Results obtained 89.7% accuracy carried on our building dataset. In addition, we compared the building detection results according to the resolution. The results showed that the lower the resolution, the lower the detection recognition.