🌍
πŸ›°οΈ
πŸ—ΊοΈ
πŸ€–
πŸ“Š
Back to Site

🌍 GeoVision

An Agent for Geospatial Analysis

Building a Geospatial Analysis Toolkit

GeoVision started as yet another hackathon project. I was among the top 100 finalists for HachWithInfy 2025 who were invited to the grand finale in Infosys Pune DC. The theme of the hackathon was "Space Tech".

One of the three problem statements was to build a "Multimodal Earth Analyst" which can help in understanding "how places change over time due to urbanization, climate shifts, infrastructure growth, or population movement". We chose this one thinking that it is the most difficult one, so least number of teams would pick it. Turned out that everyone else also thought the same and this became the most popular problem statement!

The 40 hour hackathon began with us exploring different datasets for the first 3 hours, just to realize that its better not to start from scratch, rather use some existing open source platform. After some research, we found Chat2Geo (Link). This project had everything we needed to start - a chatbot interface, Google Earth Engine integration, and a RAG (Retrieval-Augmented Generation) feature to upload custom documents.

The project was using OpenAI API which leveraged a set of tools to fetch relevant geospatial data and corresponding insights for the selected Region of Interest (ROI). We created a tool specifically for Urban Infrastructure Analysis using Copernicus Dataset, OpenStreetMap API, Overpass API and World Bank API. The tool could precisely pinpoint even the number of buildings or trees or roads that have been added or removed within an area within a specific timeframe.

The ChatBot could now easily recognize user intent and answer very specific queries like "Show me the infrastructure change in selected region of interest from May 2022 to April 2024". It didn't only provide the data, but gave actionable insights and also generated a comprehensive map highlighting the changes.

We couldn't make it to the top 3 with this one. The winning team for the same problem statement went with a multi-agent architecture and also made the chatbot multilingual plus a voice input/output feature. So they deserved it!

Key Features

πŸ—£οΈ Natural Language Interface
Ask geospatial questions in plain English and get comprehensive analysis results
πŸ›°οΈ Real-time Satellite Processing
Access Google Earth Engine's vast satellite imagery database with real-time analysis
πŸ“ Select ROI on Map
Interactively select your Region of Interest (ROI) on the map using polygon drawing tools
🧠 RAG Knowledge Base
Combine geospatial insights with your own documents using advanced RAG workflows

Analysis Types: Urban Heat Island β€’ Land Cover Change Detection β€’ Urban Infrastructure Analysis
Data Sources: Google Earth Engine β€’ Sentinel-2 (Copernicus) β€’ OpenStreetMap β€’ World Bank

Tech Stack & Integrations

Next.js 15 Typescript Google Earth Engine ArcGIS OpenAI API