Skip to main content

​
Installation Guide

DigiPathAI offers two installation modes depending on your needs. You can install just the UI for viewing whole slide images, or include the full AI pipeline with GPU support for automated segmentation.

​
System Requirements

​
Basic Requirements (UI Only)

  • Python 3.5 or higher
  • OpenSlide library installed on your system
  • 4GB RAM minimum

​
Full Pipeline Requirements

  • Python 3.5 or higher
  • OpenSlide library installed on your system
  • NVIDIA GPU with CUDA support (recommended)
  • 8GB+ GPU memory for optimal performance
  • 16GB+ system RAM
The AI segmentation pipeline requires TensorFlow GPU (version 1.14 to 1.x) and PyTorch. Ensure you have compatible CUDA drivers installed before proceeding with the full installation.

​
Installing OpenSlide

Before installing DigiPathAI, you need to install the OpenSlide library for your operating system: 
sudo apt-get install openslide-tools python3-openslide

​
Installation Options

1

Choose Your Installation Mode

Select the installation method based on your use case:

​
Option 1: UI Only (Viewer)

Install just the web-based viewer without AI capabilities. This is ideal for:
  • Viewing and exploring WSI images
  • Systems without GPU support
  • Quick deployment and minimal dependencies
pip install DigiPathAI
Installed dependencies:
  • flask==2.2.5 - Web server framework
  • openslide-python - Python bindings for OpenSlide

​
Option 2: Full AI Pipeline (with GPU)

Install the complete package including deep learning models for segmentation:
pip install "DigiPathAI[gpu]"
Additional GPU dependencies:
  • torch - PyTorch deep learning framework
  • torchvision - Computer vision utilities
  • tensorflow-gpu>=1.14,<2 - TensorFlow for legacy models
  • opencv-python - Image processing
  • imgaug - Image augmentation
  • tqdm - Progress bars
  • matplotlib - Plotting and visualization
  • scikit-learn - Machine learning utilities
  • scikit-image - Image processing algorithms
  • pydensecrf - Dense CRF post-processing
  • tifffile - TIFF file handling
  • pandas - Data manipulation
  • wget - Model downloading
Both installation methods install the same DigiPathAI package. If you start with UI-only, you can manually install GPU dependencies later to enable AI features.
2

Verify Installation

Check that DigiPathAI is properly installed:
digipathai --help
You should see the command-line help output with available options.Verify the Python module:
python -c "import DigiPathAI; print('DigiPathAI installed successfully')"
3

Test the Installation

Launch the server to ensure everything works:
# Navigate to a directory with WSI images (or an empty directory to test)
cd /path/to/slides

# Start the server
digipathai localhost 8080
Open your browser to http://localhost:8080 and verify the interface loads.
If you installed the UI-only version, the segmentation features will be disabled. You can still view all your WSI images.

​
GPU Configuration (Optional)

For optimal performance with the AI pipeline, configure your GPU settings:

​
CUDA Setup

Ensure you have CUDA 10.0 or 10.1 installed (required for TensorFlow GPU 1.x): 
# Check CUDA version
nvcc --version

# Verify GPU detection
python -c "import tensorflow as tf; print(tf.test.is_gpu_available())"

​
Memory Configuration

The segmentation pipeline automatically configures TensorFlow to allow GPU memory growth. Default settings in Segmentation.py:283: 
core_config = tf.ConfigProto()
core_config.gpu_options.allow_growth = True
You can adjust batch size and patch size to fit your GPU memory: 
from DigiPathAI.Segmentation import getSegmentation

prediction = getSegmentation(
    img_path,
    patch_size=256,   # Reduce if out of memory
    batch_size=32,    # Reduce if out of memory
)

​
Upgrading DigiPathAI

To upgrade to the latest version: 
# Upgrade UI-only installation
pip install --upgrade DigiPathAI

# Upgrade with GPU dependencies
pip install --upgrade "DigiPathAI[gpu]"

​
Troubleshooting

​
OpenSlide Not Found

If you get “OpenSlide library not found” errors:
  1. Verify OpenSlide is installed: openslide-show-properties --version
  2. On Windows, ensure OpenSlide bin directory is in PATH
  3. On Linux, try: sudo ldconfig

​
CUDA/GPU Errors

If you encounter GPU-related errors:
  1. Verify CUDA installation: nvcc --version
  2. Check TensorFlow GPU support: python -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
  3. Consider installing specific CUDA-compatible versions of PyTorch and TensorFlow

​
Import Errors

If you see import errors for specific packages: 
# Install missing dependencies manually
pip install torch torchvision opencv-python imgaug matplotlib scikit-learn scikit-image

​
Next Steps

Quick Start

Launch the server and run your first segmentation

API Reference

Explore the Python API for programmatic access

Build docs developers (and LLMs) love

Get started for freeTalk to us