Skip to content

Installation

Install CUVIS.AI and its dependencies.

Requirements

  • Python: 3.10+ (tested up to 3.13; 3.11 recommended)
  • RAM: 8GB minimum (16GB recommended; 32GB for large datasets)
  • OS: Windows / Linux / macOS
  • GPU (optional): NVIDIA + CUDA 12.8 for faster training
  • Storage: ~2GB for deps (+ space for datasets/outputs)
  1. Install uv:
# Linux/macOS
curl -LsSf https://astral.sh/uv/install.sh | sh

# Windows
powershell -c "irm https://astral.sh/uv/install.ps1 | iex"
  1. Clone and install (all extras):
git clone https://github.com/cubert-hyperspectral/cuvis-ai.git
cd cuvis-ai

uv sync --all-extras

FFmpeg (required for video pipelines)

Video nodes (VideoIterator, VideoFrameDataModule, ToVideoNode) depend on torchcodec which requires FFmpeg system libraries at runtime. Install FFmpeg before using any video functionality:

# Linux (apt)
sudo apt install ffmpeg

# Linux (conda)
conda install -c conda-forge ffmpeg

# macOS
brew install ffmpeg

# Windows (scoop) — use the *shared* build so torchcodec can find the DLLs
scoop install ffmpeg-shared
# Then add to PATH (Git Bash / MSYS2):
export PATH="/c/Users/$USER/scoop/apps/ffmpeg-shared/current/bin:$PATH"

GPU support (optional)

Check CUDA availability:

import torch
print(torch.cuda.is_available(), torch.version.cuda, torch.cuda.device_count())

Verify

Run tests:

uv run pytest tests/ -v

Skip GPU tests (CPU-only):

uv run pytest tests/ -v -m "no gpu"

Or quick import:

from cuvis_ai_core.pipeline.graph import Graph
from cuvis_ai_core.anomaly.rx_detector import RXGlobal
print("Installation successful!")

Next steps