Skip to content

Quickstart Guide

Get up and running with CUVIS.AI in 5 minutes.

Installation

First, ensure you have Python 3.10+ and uv installed:

# Clone the repository
git clone https://github.com/cubert-hyperspectral/cuvis-ai.git
cd cuvis-ai

# Install dependencies
uv sync

See the Installation Guide for detailed setup instructions.

Download Sample Data

Download the Lentils dataset from Hugging Face:

# Automated download (default: lentils dataset)
uv run download-data

# Or explicitly specify dataset
uv run download-data --dataset lentils

This downloads ~1.0 GB of real hyperspectral data to data/Lentils/.

Quick Demo: Run Pre-Trained Pipeline

Want to see CUVIS.AI in action first? Run inference with a pre-configured pipeline:

# View pipeline structure
uv run restore-pipeline --pipeline-path configs/pipeline/rx_statistical.yaml

# Run inference on sample data
uv run restore-pipeline --pipeline-path configs/pipeline/rx_statistical.yaml --cu3s-file-path data/Lentils/Demo_000.cu3s

This loads the pipeline configuration and runs anomaly detection on the sample hyperspectral cube.

Train Your Own Pipeline

Train an RX anomaly detector from scratch:

# Train RX detector
uv run python examples/rx_statistical.py

Results are saved to outputs/base_trainrun/.

What Just Happened?

  1. Loaded data - The Lentils hyperspectral dataset
  2. Built pipeline - RX statistical anomaly detector from configs/pipeline/rx_statistical.yaml
  3. Trained model - Statistical initialization on training data
  4. Saved results - Pipeline, weights, and metrics to outputs/

Use Your Trained Model

After training, restore and use your model for inference:

# Restore trained pipeline
uv run restore-pipeline --pipeline-path outputs/base_trainrun/trained_models/RX_Statistical.yaml --cu3s-file-path data/Lentils/Lentils_000.cu3s

The pipeline will load your trained weights and run inference on new data.

Next Steps

Learn the fundamentals:

Follow comprehensive tutorials:

Explore how-to guides: