Avocado Orchard Digital Twin AI Application
Avocado Orchard Digital AI is an AI-powered command center for avocado growers, designed to turn geospatial data into actionable orchard intelligence. The application uses a Cesium globe to navigate avocado-producing regions in Michoacán, Mexico, scan municipalities for orchard parcel candidates, and visualize detected parcels with GPS boundaries, estimated hectares, tree counts, NDVI, stress levels, and confidence scores. The backend is built with FastAPI and supports orchard detection, PostgreSQL-based archive persistence, company orchard networks, financial impact analysis, and Vision/3D analysis. A live AMD MI300X vLLM endpoint running Qwen3-32B powers the AI advisor for natural-language commands, workflow guidance, and grower recommendations. Users can scan an area, save detected orchards to an archive, compare municipalities, analyze financial risk, generate analytics summaries, and create 3D digital twin views for selected orchard parcels. The goal is to help growers make faster, data-informed decisions about monitoring, irrigation, stress risk, production planning, and orchard management.
Category tags:.png&w=96&q=75){kind=small}
"Impressive — solid solo build with real GPU-powered AI, actual agricultural value, and a complete end-to-end workflow. Application of Technology 4/5 Solid tech stack: Cesium, FastAPI, vLLM, Qwen3-32B on AMD MI300X, Three.js for 3D twins, PostgreSQL. Good integration of multiple technologies. Business Value 4/5 Real agricultural use case - avocado industry in Michoacán is a $billion market. The workflow (scan → archive → analyze → 3D twin) addresses actual grower pain points. Could stronger justify ROI. Presentation 4/5 Comprehensive README, clear architecture diagrams, well-documented API. Couldn't verify live demo but the docs are thorough. Originality 3/5 Geospatial + agriculture AI exists, but the specific combo (avocado orchards + Cesium + digital twins + AMD GPU inference) is relatively unique. Strengths ✅ Impressive end-to-end workflow: satellite scanning → parcel detection → archive → analytics → 3D digital twin ✅ Live AMD MI300X + Qwen3-32B inference is a solid AI showcase ✅ Practical use case with clear value proposition ✅ Good fallback logic for when dependencies unavailable Weaknesses ⚠️ Digital twin generation appears simulated rather than true computer vision ⚠️ Business model / monetization not clearly addressed "
Sanem Avcil
For Innovators & Creators