Space to Policy: Scalable Brick Kiln Detection and Automatic Compliance Monitoring with Geospatial Data
Object Detection
Deep Learning
Satellite Imagery
Paper accepted in ACM Journal on Computing and Sustainable Societies
Figure: A few samples of predicted brick kilns of each brick kiln category from our model. The first, second, and third rows correspond to CFCBK, FCBK, and Zigzag, respectively.
- Objective: Built a scalable machine learning pipeline to detect and classify 30,638 brick kilns across five states in India using Planet Labs satellite imagery.
- Context: Brick kilns contribute 8–14% of India’s air pollution — accurate detection is vital for emission inventories and environmental governance.
- Technical Approach:
- Trained YOLO-OBB models for oriented bounding box detection.
- Developed a custom geo-annotation tool using Leafmap for labeling and visualization.
- Implemented iterative hand-validation improving detection precision to 82%.
- Trained YOLO-OBB models for oriented bounding box detection.
- Dataset: Created the largest open brick kiln dataset (30k+ validated samples across 520,000 km²).
- Validation: Strong real-world correlation (r = 0.94) with official UPPCB 2023 survey.
- Impact: Enabled automated compliance analysis and policy integration, bridging the gap between AI, satellite imagery, and governance.
To know more check: Paper