Welcome to U-FISH Documentationļƒ

PyPI version

U-FISH šŸŽ£ is an advanced FISH spot calling algorithm based on deep learning. The ā€œUā€ in U-FISH represents both the U-Net architecture and the Unified output of enhanced images, underpinning our design philosophy.

U-FISH Overview

Key Featuresļƒ

  • Diverse dataset: 4000+ images with approximately 1.6 million targets from seven sources

  • Small model: State-of-the-art performance with only 160k parameters (ONNX file size: 680kB)

  • 3D support: Detection of FISH spots in 3D images

  • Scalability: Support for large-scale data storage formats (OME-Zarr and N5)

  • User-friendly interface: API, CLI, Napari plugin, and web application

Quick Startļƒ

Installationļƒ

pip install ufish

For GPU inference support:

pip install onnxruntime-gpu

Basic Usageļƒ

from skimage import io
from ufish.api import UFish

# Initialize U-FISH
ufish = UFish()
ufish.load_weights()

# Predict spots
img = io.imread("path/to/image.tiff")
pred_spots, enh_img = ufish.predict(img)

Contents:

Indices and tablesļƒ