buffalo-udder-seg: Thermal Infrared Images of Italian Mediterranean Buffalo Udders with Pixel-Level Segmentation Masks

buffalo-udder-seg Summary Open dataset of long-wave infrared (LWIR) thermal images of Italian Mediterranean buffalo udders with expert-verified pixel-level segmentation masks. Images were acquired in situ during robotic milking and curated to support computer-vision research on udder region detection/segmentation in thermography, a prerequisite for non-invasive estimation of udder skin surface temperature (USST) and early mastitis monitoring. Contents This release contains normalized (min-max temperature normalization) single-channel images (PNG, 8-bit) and binary masks (PNG, 8-bit; udder = 255, background = 0) with predefined train/val/test (70/20/10) splits. The total number of image-mask pairs is 2,148, at a resolution of 640 x 480 pixels. Acquisition and calibration Sensor: FLIR A700 (LWIR, 640 x 480, <30 mK NETD) Time range: 2022-04-02 to 2023-06-11 (YYYY-MM-DD) Location: Buffalo farm in Cancello e Arnone, in Caserta, southern Italy, with about 1000 animals Setup: Fixed rear-facing camera on the robotic milking unit; approximate 1 m distance; emissivity 0.98; focus locked Environment: Ambient temperature and relative humidity logged via a nearby weather station were used for temperature correction before normalization File structure images/├─ train/│ ├─ <uuid>.png│ ├─ <uuid>.png│ └─ …├─ val/│ ├─ <uuid>.png│ ├─ <uuid>.png│ └─ …└─ test/ ├─ <uuid>.png ├─ <uuid>.png └─ …masks/├─ train/│ ├─ <uuid>.png│ ├─ <uuid>.png│ └─ …├─ val/│ ├─ <uuid>.png│ ├─ <uuid>.png│ └─ …└─ test/ ├─ <uuid>.png ├─ <uuid>.png └─ … Annotation protocol Two veterinary experts independently annotated udder polygons in a data labeling platform following a shared guideline. Inter-observer disagreements were reviewed and resolved by consensus. Intended use Benchmarking segmentation models on thermal livestock imagery Pretraining/transfer for agricultural/veterinary thermal vision Research on precision livestock farming and early mastitis indicators via downstream USST extraction Ethics and privacy Thermal imaging was performed during routine husbandry without the need for additional restraint beyond standard robotic milking. Frames show the udder region only and contain no personally identifiable information. The farm owner provided permission for the use and release of data for research purposes. Acknowledgements This work was carried out within the Agritech National Research Centre and received funding from the European Union's "NextGenerationEU" under the Italian PNRR (CN00000022), and from the European Union's “Future Artificial Intelligence Research (FAIR)” under the Italian PNRR (E63C22002150007). Additional information For any additional information about the collection procedure, dataset statistics, and more, please refer to this data article. Citation If you use this dataset, please cite this record and the data article describing this dataset as (BibTeX format):@article{FONISTO2026112587,title = {Thermal Infrared Italian Mediterranean Buffalo Udder Open Dataset With Expert-Verified Segmentation Masks},journal = {Data in Brief},pages = {112587},year = {2026},issn = {2352-3409},doi = {https://doi.org/10.1016/j.dib.2026.112587},url = {https://www.sciencedirect.com/science/article/pii/S235234092600140X},author = {Mattia Fonisto and Maria Teresa Verde and Francesco Bonavolontà and Annalisa Liccardo and Roberta Matera and Matteo Santinello and Flora Amato},keywords = {Thermal Imaging, Veterinary Imaging, Buffalo Udder, Mastitis Screening, Semantic Segmentation, Computer Vision},abstract = {This data article describes an open dataset of thermal images of buffalo udders curated for pixel-level segmentation, the first publicly accessible open dataset on the topic. Images were acquired in situ at a commercial farm in southern Italy over more than 1 year during routine robotic milking, using a fixed, rear-facing long-wave infrared camera (640×480 resolution) positioned at approximately 1 m, with emissivity set to 0.98. Environmental measurements (ambient temperature and relative humidity) from a nearby station were recorded at capture time and used for temperature compensation before min–max conversion to single-channel 8-bit PNGs. The release contains 2,148 image–mask pairs organised in predefined train/validation/test splits. Two veterinarians produced udder masks with polygonal annotation following a shared protocol and resolving disagreements by consensus. The dataset is intended to support reproducible research on udder-region segmentation, a necessary preprocessing step before extracting udder skin surface temperature (USST) in mastitis-oriented studies of dairy animals but not limited to only mastitis. Beyond udder health, thermal computer vision tasks in veterinary science include behaviour monitoring and localisation of anatomical regions for temperature tracking. Labelled thermal segmentation data can also serve as source material for pretraining and transfer to such applications.}}