AI-NICC: AI-Based Framework for Non-Invasive Classification of Cats’ Health
- Source
- DataCite
- Published
- 2026
- License
- Creative Commons Attribution 4.0 International
- Themes
- Animal Welfare, Machine Learning & AI
About this dataset
The AI-NICC Dataset is a rigorously standardized and augmented derivative of the Thermal Imaging Cats’ Dataset (TICD), engineered specifically for reliable, non-invasive cats' health classification ("Healthy" vs. "Sick"). To eliminate ambient environmental thermal noise, this dataset strictly isolates 74 unique cats captured in controlled indoor clinical settings. It utilizes a strict Patient-Level Split (50 Train / 12 Validation / 12 Testing) to prevent data leakage and guarantees Multimodal Evaluation Parity by pairing image directories with synchronized 11-feature tabular CSVs. Furthermore, the dataset addresses clinical prevalence imbalance through Thermal-Safe Deterministic Augmentation dynamically oversampling the training set to 7,588 images using spatial transformations and simulated sensor noise, while strictly prohibiting photometric distortions to preserve raw diagnostic...
How the data was collected
Furthermore, the dataset addresses clinical prevalence imbalance through Thermal-Safe Deterministic Augmentation dynamically oversampling the training set to 7,588 images using spatial transformations and simulated sensor noise, while strictly prohibiting photometric distortions to preserve raw diagnostic temperature data.