rapidsai/cucim
cuCIM - RAPIDS GPU-accelerated image processing library
What it solves
cuCIM addresses the performance bottlenecks associated with processing large, multidimensional images. It is specifically designed for high-resolution data used in fields like biomedical imaging, geospatial analysis, remote sensing, and material science, where traditional CPU-based processing is often too slow.
How it works
cuCIM leverages NVIDIA GPUs to accelerate computer vision and image processing primitives. It provides a Pythonic interface and an API that matches Openslide, making it easier for users to integrate into existing workflows. The library supports a wide range of complex image formats, including Aperio ScanScope Virtual Slide (SVS) and various Tiled Multi-resolution RGB TIFF files with multiple compression schemes (JPEG, JPEG2000, LZW, and Deflate).
Who it’s for
This tool is for researchers and developers in life sciences, geospatial analysis, and remote sensing who work with massive n-dimensional image datasets and require GPU acceleration to speed up their analysis pipelines.
Highlights
- GPU Acceleration: Uses Graphics Processing Units to speed up image processing and computer vision tasks.
- Broad Format Support: Handles SVS, Philips TIFF, and generic tiled TIFFs with various compression types.
- Openslide Compatibility: Offers an API that matches Openslide for a straightforward transition.
- Multidimensional Support: Specifically optimized for large n-dimensional tag image file format (TIFF) files.