11 results for “topic:geospatial-ai”
Geoff (GEOspatial Fact Finder) is a geospatial AI that turns natural language into spatial queries and displays results on a web map
Winner of FIAP'S Global Solution 2025.1 Challenge. This repository contains the architecture for a multi-agent system where five autonomous "Guardians" work in synergy to predict, manage, and respond to complex events like wildfires, epidemics, and infrastructure failures.
TN-Epic is the world’s first State-Scale Experience OS, transforming Tamil Nadu into a playable AR Meta-Game. By turning monuments into levels and merchants into NPCs, we solve the 80% revenue leakage and ₹1,200Cr sanitation crisis. Built with ARCore, YOLOv8, and Java Spring Boot, it gamifies civic duty for 286M annual tourists.
साहित्याच्या पुणेरी पाट्या - Where Marathi literature meets AI. Explore 47+ locations and 60+ quotes across Pune through an interactive 3D map powered by vector search and iconic 'Puneri Patya' design.
Geospatial deep learning pipeline for detecting illegal mining structures and machinery in satellite imagery using YOLO and spatial post-processing.
FRA Atlas AI is an intelligent forest rights digitization and monitoring system that automates document verification via OCR, classifies land using CNN & ISRO Bhuvan data, and powers a LangChain-based Decision Support System for transparent, data-driven governance.
AI-powered wildfire damage detection using IBM/NASA Prithivi-EO-2.0 foundation models and hybrid spectral analysis for high-precision burn mapping.
AI-Powered Climate Vulnerability Assessment for Transportation Infrastructure using Google's Satellite Embeddings
A comprehensive, end-to-end geospatial pipeline for wildfire damage assessment. Integrates Google Earth Engine (GEE) data harvesting with NASA-IBM Prithvi EO 2.0 (Vision Transformer) fine-tuning to achieve a +50.58% Macro F1 improvement in burn scar severity classification.
PineGuard: A geospatial AI platform for Pine Wilt Disease (PWD) detection. Integrates 40-year Sentinel-2 spectral analysis, land-cover masking, and regression modeling to identify infection epicenters and project 20-year strategic risk.
Binary classification experiments to interpret Google AlphaEarth Foundation embeddings across ESA WorldCover land cover classes. Part of the study "What on Earth is AlphaEarth?" — 130,000+ experiments using Random Forest, XGBoost, LightGBM and progressive ablation across 64 embedding dimensions.