Construction of human detection model in indoor scene based on YOLO V4
Xiaoman Liu, Xiaohui Wang, Lv Qiongshuai · 2024
With the development of artificial intelligence technology, indoor human detection technology under computer vision is widely used in video surveillance, smart home, human-computer interaction and other fields. In the face of increasingly complex and diverse application scenarios, the traditional statistical classification algorithm has some problems, such as insufficient environmental adaptability, weak small target detection ability and high computational complexity, which makes it difficult to adapt to the complexity and variability of indoor scenes. In this regard, based on the actual project application requirements, this paper will propose a set of human detection model based on YOLO v4 algorithm, which is used to achieve efficient and accurate human detection in indoor scenes. Practice has proved that the average detection precision of human detection model in indoor scene based on YOLO v4 algorithm on each test data set is 96.61%, and the detection speed is 63.1 frame/s. Compared with the traditional statistical classification algorithm, the detection efficiency and detection effect are obviously improved, which can meet the practical application requirements.