TIDS: A Thermal Imaging Dataset for Subclinical Mastitis in Dairy Sheep

Georgios Botsoglou, Marios Lysitsas, Dimitris Dimitriadis, Constantina N. Tsokana, George Valiakos, Grigorios Tsoumakas

Abstract: Subclinical mastitis (SCM) in dairy sheep is a significant challenge in the agricultural and veterinary sectors, leading to substantial financial losses for farmers and negatively impacting overall dairy sheep productivity. It often goes unnoticed due to the absence of clinical signs, making early diagnosis particularly difficult. However, early identification of SCM is critical, as it allows for timely intervention that can prevent disease progression, reduce economic losses, and minimize the need for costly treatments. This work focuses on detecting SCM in dairy sheep using a non-invasive thermal imaging approach. A major limitation in this field is the lack of available datasets, as acquiring such data presents several challenges. Capturing clear thermal images of dairy sheep udders is hindered by factors such as animal movement, environmental conditions, and variability in breed, health, and udder size. Furthermore, large, diverse sample sizes are required, making data acquisition resource-intensive. Ethical concerns regarding animal welfare and the high cost of thermal imaging equipment add to the complexity. These challenges hinder the use of data-driven techniques, such as deep learning models, which require large datasets. In this paper, our contributions are two-fold: first, we introduce a novel dataset, TIDS, along with an explanatory analysis supported by domain expertise. Second, we apply deep learning models to detect SCM in dairy sheep and provide a comprehensive methodology, marking a novel approach in this area.

External IDs:dblp:conf/pkdd/BotsoglouLDTVT25