Dataset of Low-Resolution Thermal Images for Water Stress Classification in Blueberry

This dataset contains raw environmental data and low-resolution thermal images (32x24 pixels) captured using an MLX90640 sensor. It was built to train a Multilayer Perceptron (MLP) neural network capable of classifying water stress in 'Biloxi' blueberry plants (Vaccinium corymbosum L.). The data was collected over 55 continuous days (June - July 2025) inside a micro-greenhouse in Cundinamarca, Colombia, utilizing a custom low-cost IoT architecture based on an ESP32 microcontroller.The repository includes the original tabular data (thermal matrices and environmental variables) and the Python script used to calculate the empirical Crop Water Stress Index (CWSI), which establishes the baseline to label the images into two categories: "healthy" (recovery phase) and "stressed" (water deficit phase).