Anveshak is a powerful tool for researchers and scientists to explore and analyze large-scale planetary image datasets. It allows users to ingest vast amounts of image tiles, process them into a searchable index, and then perform sophisticated similarity searches to find geographical features of interest. The application has a machine learning-powered backend that uses computer vision to identify and compare features within the images, and a frontend with an intuitive map-based interface.
The backend is built with FastAPI and uses libraries such as timm for feature extraction, faiss for efficient similarity search, and rasterio for geospatial data processing. The frontend is a single-page application built with Svelte, utilizing Leaflet.js for interactive maps. The project also includes a data ingestion pipeline handled by a separate Python script, fetch_footprints.py, which fetches metadata from NASA's Trek APIs.
- Python 3.12 or newer
- Node.js and npm
- uv (a Python dependency management tool)
-
Extract Data: Extract the
data.rarfile into the current directory. -
Backend Setup:
- Run
uv syncto install backend dependencies. - Run
./run.shto start the backend server.
- Run
-
Frontend Setup:
- Navigate to the frontend directory:
cd frontend - Install dependencies:
npm install - Start the development server:
npm run dev
- Navigate to the frontend directory: