SquareTrack: A Novel Approach for Tracking Pigs in Commercial Swine Farming

Monitoring animal behavior is a very important activity in commercial farming, especially in pig farming. For example, aggressiveness is an undesirable type of behavior in a farm setting, requiring constant monitoring by the farmer, being prone to errors, and time consuming which can be reduced through automation using tracking algorithms. Computer vision and artificial intelligence, employing neural networks, allow for the implementation of techniques to track animals in videos. SquareTrack is a technique developed for this purpose and has been compared with DeepSort, ByteTrack, and BotSort. The superior performance of SquareTrack on the MOTA and HOTA_ALPHA metrics was 0.66 and 0.72, respectively, for tracking pigs using a dataset obtained from a commercial farm.