BHOOMRA

Inspiration

India’s farms are small, diverse, and often hidden under monsoon clouds, making crop monitoring harder than anywhere else. We were inspired to build a solution that combines frontier AI and satellites to give farmers and policymakers real intelligence.

What it does

BHOOMRA will identify crop types across India using multi-modal satellite data (SAR + optical). It will produce district-level crop distribution maps and dashboards that can guide farmers, governments, and agribusinesses.

How we plan to build it

Collect data from Sentinel-1, Sentinel-2, EOS-4, ResourceSAT-2/2A.
Preprocess with cloud masking, temporal stacking, field boundary extraction.
Fine-tune a geospatial foundation model using AgriFieldNet India dataset.
Validate across states (UP, Bihar, Rajasthan, Odisha).
Deploy prototype on Kaggle with a dashboard interface.

Challenges we anticipate

Handling cloud cover and missing data in optical imagery. Making models generalize across diverse Indian farm practices. Balancing accuracy and explainability for real adoption.

Accomplishments we’re proud of

So far, we’ve developed a strong research-backed proposal that combines the best of SAR + Optical + Foundation Models, tailored for India.

What we expect to learn

How foundation models perform on fragmented Indian agriculture. The trade-offs between model complexity, accuracy, and explainability. Practical deployment challenges of AI on real satellite data.

What’s next for BHOOMRA

Begin implementation on Kaggle once the proposal is accepted. Build district-level dashboards. Extend coverage nationwide and explore partnerships with government and agri-startups.