Integration of MobileNet-SSD and Isolation Forest as a Prototype of Web-Based Chicken Movement Anomaly Detector

In broiler farming, animal welfare is one aspect the breeders must heed. Two guiding principles of animal welfare include freedom from hunger or thirst and freedom from pain, injury, or disease. However, without further surveillance, continuous feeding activity on broilers can result in overweight broilers. These overweight broilers have difficulty moving to obtain nutrients around them, resulting in their sickness and death. The condition violates two of the abovementioned animal welfare principles. The rare movement of broilers is considered an anomaly. The availability of a web application that detects such broiler movement anomaly is beneficial. In this study, a web prototype to detect broiler movement anomaly is developed. In the prototype backend, the Python Flask framework integrates the MobileNet-SSD, centroid tracking, and Isolation Forest algorithm from previous studies. The integration involves connecting the centroid tracking result in one study with another study's Isolation Forest detection algorithm. The integration, therefore, reduces time complexity due to the substitution of Simple Online and Realtime Tracking with Euclidean centroid tracker. User interfaces are also developed in the prototype, allowing users to quickly input broiler surveillance video and receive the processing results. The main developed user interfaces are the homepage, page to insert video, and page displaying process results. Functionality testing has shown that the web prototype features functioned as expected and can return the expected result. Unlike prior studies, this study introduces the first prototype for detecting broiler movement difficulty, with a modularity that facilitates easy substitution of modules in further research.