Detection of unwanted objects, alives and situations in the food production sector by machine learning methods

Depending on the population growth rate of Turkey, there is a rapid development in the food production and distribution sector. Population changes have caused companies that carry out food production to increase both the number and production capacities. This capacity increase has led to the employment of more people. However, inspections of undesirable situations in food production areas are having difficulty coping with this rapid growth. These inspections in food production areas are usually carried out manually or not at all, which threatens human health. Artificial intelligence-based systems have started to solve this problem more effectively with the development of technology. With this thesis, it is aimed to develop an object detection system for use in the food production and distribution sector. The developed system detects whether the personnel are wearing gloves, masks and bonnet equipment through the camera images. At the same time, different model files were used to detect unwanted creatures in the food production environment. The YOLOv5 algorithm used in the study stands out with its high accuracy rates and fast detection ability. Glove-Mask and Bonneted-Boneless models can successfully determine whether the personnel are using hygiene equipment. In addition, the Cat-Dog detection model helps to minimize potential risks in terms of hygiene. The developed Python-based interface allows users to easily manage the system. The sections listing the addition of cameras, live image monitoring and detection of abnormal conditions make the system user-friendly. The achievements obtained make the object detection system developed an effective solution in industrial applications.