Israel is a major global migration route for many species of fish-eating birds, such as pelicans and cormorants. This migration carries high costs for aquaculture in fish predation and the spread of disease in the fish ponds. The aim of this study was to develop a computerised automatic system to detect and deter fish-eating birds from fish ponds. We developed a system that detects pelicans and sends a small boat to deter them. The system comprises a radio-controlled boat (1 m long), a stationary video camera, a computer, and a detection algorithm. The camera (Dahua camera, DH-IPC-HFW4830E-S) was connected to the computer via an Ethernet cable to collect video sequences of a focal pond. The detection algorithm is based on a convolutional neural network, and classifies different regions as containing pelicans, or not. Data were recorded over the three months before the migration season. The data contained videos of between two and 50 pelicans per video. The size of the test fish pond was 105 m2 with a maximum range of 600 m to the pelicans as they floated. The detection complexity varies according to air clarity, cloud cover, and target proximity. Pelicans were detected on clear days, cloudy days, and with different distances to the camera. The detection accuracy was 98%. The accuracy was calculated based on a dataset that was built for this project. The system is currently being tested and is on the way to become fully automatic.