cuCIM

Type: Technology Tags: CUDA, NVIDIA, GPU, Image Processing, Computer Vision, RAPIDS, scikit-image, Open Source Related: NVIDIA-RAPIDS, cuDF, cuML, NVIDIA-DALI, CV-CUDA, NPP Sources: NVIDIA official documentation (RAPIDS), https://docs.nvidia.com/rapids/index.html, https://developer.nvidia.com/topics/ai/data-science/cuda-x-data-science-libraries, https://docs.rapids.ai/api/cucim/ Last Updated: 2026-04-30

Summary

cuCIM (CUDA Clara IMage) is a GPU-accelerated image processing library that mirrors scikit-image’s API for GPU execution, and incorporates OpenSlide for whole-slide image loading. Part of NVIDIA-RAPIDS, it accelerates image manipulation and computer vision operations for scientific, medical, and industrial imaging workflows where scikit-image is too slow for large image volumes.

Detail

Purpose

Scientific and medical imaging workflows often process thousands of high-resolution images using scikit-image operations (filtering, morphology, feature extraction). cuCIM provides GPU-accelerated equivalents with a compatible API, enabling existing scikit-image workflows to run dramatically faster without major code changes.

Key Features

  • GPU-accelerated scikit-image compatible API for image processing
  • OpenSlide integration for loading whole-slide images (WSI) used in digital pathology
  • Image manipulation: filters, morphological operations, feature extraction, transforms
  • Support for multi-dimensional image arrays (2D, 3D, N-D)
  • Integration with RAPIDS cuDF and cuML for end-to-end GPU pipelines
  • Part of the RAPIDS ecosystem — compatible with other RAPIDS libraries
  • Python API

Use Cases

  • Digital pathology: whole-slide image analysis and classification
  • Medical imaging: MRI/CT slice preprocessing and feature extraction
  • Scientific microscopy image processing
  • Satellite and aerial imagery analysis
  • Industrial quality control image processing
  • Training data augmentation for computer vision models

Hardware Requirements

  • NVIDIA GPU, Pascal or newer (Volta+ recommended)
  • CUDA 11.x or 12.x
  • Linux (primary supported OS)
  • Part of RAPIDS ecosystem

Language Bindings

  • Python (primary API, scikit-image compatible)

Connections

  • NVIDIA-RAPIDS — cuCIM is the scientific image-processing library in the RAPIDS/CUDA-X data science family
  • cuDF — cuCIM integrates with cuDF for metadata and tabular data associated with images
  • cuML — cuML models consume features extracted by cuCIM for classification/detection
  • NVIDIA-DALI — DALI handles training data loading; cuCIM handles scientific image processing
  • CV-CUDA — CV-CUDA focuses on inference preprocessing; cuCIM mirrors scikit-image for scientific use
  • NPP — NPP provides lower-level image primitives; cuCIM provides higher-level scikit-image API

Resources