Inspiration

India’s farmlands are small, fragmented, and diverse, making crop identification a persistent challenge. Traditional methods often fail to capture this complexity, leading to inefficiencies in yield prediction, resource allocation, and farmer support. We wanted to harness the power of AI and satellite imagery to create a solution that works for Indian agriculture and empowers both policymakers and farmers.

What it does

CropVision classifies crops across districts in India using multi-source satellite data (SAR from Sentinel-1, multispectral from Sentinel-2, and optical from ResourceSAT). It produces accurate crop maps that can be visualized interactively, helping stakeholders track crop types, plan resources, and respond quickly to agricultural needs.

How we built it

We used geospatial foundation models fine-tuned with Indian datasets, along with CNN and transformer-based architectures for classification. By fusing SAR and optical data, our system overcomes cloud cover challenges. The pipeline processes satellite imagery, trains the model, and outputs clear, interpretable crop classification maps.

Challenges we ran into

Handling massive geospatial datasets and preprocessing satellite imagery. Balancing accuracy with generalization across diverse crop types. Integrating SAR and optical data effectively. Limited ground-truth datasets for Indian farmlands.

Accomplishments that we're proud of

Building a working prototype that classifies multiple crops with good accuracy,successfully fusing SAR and optical imagery for reliable results in monsoon conditions and designing an interactive visualization tool to make results accessible.

What we learned

We learned how powerful foundation models and multi modal data fusion can be for solving real-world agricultural challenges. We also gained experience in preprocessing complex geospatial datasets and applying AI in an India-centric context.

What's next for CropVision

Expanding to more rural areas and crop types across India. Collaborating with agricultural departments and NGOs for ground validation. Turning the prototype into a scalable decision-support tool for policymakers and farmers. Exploring integration with mobile-based advisory systems for direct farmer benefits.