DIIP-Dataset for Pig Interaction Identification with an Enriched Environment

Identification of changes in pig behavior or interaction such as playing, sniffing, chewing, lying, or aggression is important for taking the necessary action if needed. Manual identification of pig behavior by human observers is not possible because it requires continuous monitoring. It is, therefore, essential to develop an automated method that quantifies pig behavior. The proposed dataset consists of 7187 images with corresponding annotated files in text and XML format of different behavior classes, which focus mainly on the interaction of pigs with elements of their environment. This annotated dataset can be used to train the AI Algorithm for the development of Deep Learning computer vision models for pig interaction identification.

Dataset is published with IEEE DataPort. Please follow the DOI to access the dataset.

This research was EU Horizon 2020 funded.