EarthLake: Model Management for Earth Observation Foundation Models

Overview

EarthLake is a model lake system for Earth Observation (EO) foundation models.
It enables model discovery, evaluation, and operation in a unified workflow.

Features

• 🔎 Model Discovery via natural language (REMSA)
• 📊 Benchmarking & Evaluating of EO models
• 🚀 Operation & Inference on new imagery
• 🗂️ Model Registry with structured metadata

Architecture

REMSA Chat Tool

Benchmarking

Getting Started

1. run cp .env.example .env
2. Set your OpenAI Api Key in .env
3. Run setup with correct profile ( 'gpu' for cuda, else 'cpu', including mps )
4. docker compose --profile cpu up
5. go to http://localhost:5173

Acknowledgments

The original implementation of REMSA: https://github.com/be-chen/REMSA.git