Crop Burn Detection Labeled is a collection of paired RGB and SWIR Sentinel-2 satellite image tiles annotated for crop residue burning. It covers 68 agricultural districts in northern India during the paddy (October–November 2025) and wheat (March–May 2025) stubble burning seasons. The dataset was created by munish0838 to train and benchmark vision models for real-time burn detection.
Use Cases
- Training supervised image classification models based on per-tile burn annotations.
- Benchmarking remote sensing models for agricultural event detection based on paired RGB and SWIR imagery.
- Analyzing spatial patterns of crop residue burning across northern India based on district-level coverage.
- Developing real-time monitoring systems for agricultural air pollution based on seasonal satellite imagery.
Strengths
- Provides per-tile burn annotations for supervised learning.
- Covers two distinct agricultural seasons (paddy and wheat) across 68 districts.
- Includes paired RGB and SWIR spectral bands from Sentinel-2 imagery.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Freshness should be verified; last updated metadata is from 2026-05-04.
Provenance
- Source
- munish0838 on Hugging Face.
- Collection Method
- Built from Sentinel-2 satellite imagery with manual or automated annotations.
- Time Range
- October–November 2025 and March–May 2025.
- Freshness
- Last updated 2026-05-04 10:28:48.
- Geography
- Agricultural districts of northern India.