A High-Resolution Proximal Multispectral Dataset for Early Detection of Brown Spot (Bipolaris oryzae) in Tropical Rice Crops

This dataset provides high-resolution proximal multispectral imagery specifically designed for the detection and classification of Brown Spot disease (Bipolaris oryzae) in rice crops (Oryza sativa L.). The data were acquired in the tropical savannas of Casanare, Colombia (Yopal and Aguazul municipalities), representing one of the most productive rice regions in the country.

1. Data Acquisition and Sensor Specifications: The imagery was captured using a DJI P4 Multispectral (P4M) platform. The system records six discrete spectral channels: one RGB (visible) and five monochromatic bands centered at Blue (450 nm), Green (560 nm), Red (650 nm), Red-Edge (730 nm), and Near-Infrared (840 nm). To resolve early-stage necrotic lesions (often smaller than 1.5 mm), a proximal sensing strategy was implemented, maintaining a sensor height of 30–45 cm from the canopy with a 45° oblique perspective.

2. Dataset Composition: The dataset consists of 2,772 multispectral sets (totaling over 13,000 individual TIFF files) categorized into two primary classes:

Healthy: Asymptomatic rice leaves confirmed under optimal nutrient conditions.

Diseased (Brown Spot): Leaves exhibiting characteristic oval lesions with necrotic centers and chlorotic halos, ranging from early to advanced infection stages.

3. Ground Truth Validation: Class labels were assigned through a rigorous \textit{in-situ} inspection protocol. Ground truth was validated by expert agronomists from the TICTROPICO research group (Unitropico) and local technical assistants, following the phytosanitary monitoring standards for Colombian tropical rice.
4. Potential Applications: This dataset is optimized for training and benchmarking Deep Learning architectures (e.g., CNNs, ConvNeXt, Vision Transformers) and for the development of high-dimensional Bio-Spectral Tensors. It serves as a benchmark for precision phytopathology and autonomous crop health monitoring in tropical agricultural ecosystems.