contact me: feimos@mail.ustc.edu.cn
Crystal – A comprehensive, intelligent and efficient medical 3D visualization engine
Crystal features a powerful built-in rendering engine that supports rendering various data types, including but not limited to mesh data, volumetric data, point clouds, Gaussian kernels, and implicit neural field data. It offers multiple rendering modes such as volumetric path tracing, mesh-volume hybrid rendering, neural radiance field rendering and point cloud rendering. A notable feature of the Crystal engine is its support for cinematic-quality rendering effects across various visualization modes.
pycrystal
· PyCrystal
Crystal research domain
· Realistic Rendering
· Geometric Analysis and Processing
· Medical imaging analysis
· Neural rendering and 3D reconstruction
· Ultra large scale data
· Mixed reality and surgical simulation
· Commercial game engine
· Data analysis
covered areas and characteristics
In the Crystal project, the eight specialized domains include: rendering algorithms and engines, neural rendering and 3D reconstruction, geometry processing and optimization, medical image analysis and processing, integration with commercial game engines, mixed reality and surgical simulation, data analytics and server management, and large-scale data visualization. Core areas within Crystal encompass medical image analysis and processing, rendering algorithms and engines, and geometry processing and optimization. Medical image analysis covers techniques such as denoising, information enhancement, multi-modal registration, and semantic segmentation. Geometry processing involves converting semantic segmentation results of volumetric data into mesh data and performing smoothing optimization. The rendering engine renders various data types to desktop displays. Neural rendering and 3D reconstruction include tasks like neural denoising, neural rendering, and inverse rendering, primarily serving academic research. We provide data generation and algorithmic support for such research. Commercial game engine integration focuses on cross-platform compatibility, and as mixed reality technology emerges as a future research trend, Crystal also offers corresponding support.
News !
[2025.0226] The Crystal engine supports dragging volume data files, allowing only mask changes to be made. Starting to support single frame multi spp rendering, 4196 spp high-definition and low-noise images can be obtained within 20 seconds.
[2025.0213] Python support has been provided for the renderer, allowing the rendering functionality to be directly called from Python scripts, and enabling the direct import of rendering output arrays into Python programs!
# Deploying pip installation, expected to be available in April
pip install pystal
>>>import pystal as ctl
>>>ctl.load_data("scene.txt","datamapper.txt")
>>>ctl.set_spp(1024)
>>>ctl.rendering()[2025.0116] We have generated compatible formats for neural rendering and 3D reconstruction techniques.
Software startup interface and rendering example interface
Crystal adopts a streamlined startup interface similar to Mitsuba, allowing scenes to be loaded via files through this interface. Nearly all parameters and modes can be predefined and saved in advance. Additionally, Crystal supports loading parameters via command line and initiating rendering to output images.
Multiple display effects and auxiliary functions
Crystal supports multiple visualization effects and can handle complex multidimensional transformation functions for debugging and analysis purposes.
Multi level efficient acceleration structure
Crystal supports various acceleration structures including octrees, voxel grid partitioning, KD-trees, and BVH trees to accelerate volumetric rendering, photon mapping, and surface model rendering speeds.
Various examples
Crystal provides a rich set of example scenes that users can easily access directly from the startup interface.
Datasets
Volumetric Datasets:
OpenOrganelle: https://openorganelle.janelia.org/datasets
Osirix: https://www.osirix-viewer.com/resources/dicom-image-library/
EMDB: https://www.ebi.ac.uk/emdb/
Medical Decathlon: http://medicaldecathlon.com/
CBCTData: https://www.kaggle.com/datasets/imaginar2t/cbctdata?resource=download
Stanford volume data: https://graphics.stanford.edu/data/voldata/voldata.html
3DSlicer SampleData: https://www.slicer.org/wiki/SampleData
Open Scivis Datasets: https://klacansky.com/open-scivis-datasets/
IDA: https://ida.loni.usc.edu/login.jsp
linhandev: https://github.com/linhandev/dataset
HDR Panorama Environment Map
laval-indoor-hdr-dataset
free-hdri-environments
hdri-hub
Mesh Datasets
https://sites.cc.gatech.edu/projects/large_models/
https://graphics.stanford.edu/data/3Dscanrep/
https://casual-effects.com/data/
https://pbrt.org/scenes-v3
https://benedikt-bitterli.me/resources/
https://www.3drender.com/challenges/
https://free3d.com/