UC Davis researchers needed a practical way to create convincing materials for scientific datasets under multiple lighting conditions, while also overcoming the long runtimes and memory constraints involved in training and refining custom diffusion control models on typical research hardware.
Reported impact
The pipeline enabled UC Davis to synthesize realistic materials for scientific visualization while significantly reducing training and inference time, making rapid iteration on model design and material generation practical in a research setting.
UC Davis VIDI built a prompt-driven generative AI pipeline for scientific visualization that combines Stable Diffusion with a custom-trained ControlNet model to author materials and textures for rendering. The workflow ingests text prompts, sphere map examples, lighting-direction controls, and training grids rendered from multiple material models with spot lighting at varied orientations. Stable Diffusion provides the pretrained visual prior, while the custom ControlNet enforces pixel-level lighting-direction constraints so the system can generate coherent sphere-map grids under predefined illumination. Training and inference run on Intel Data Center GPU Max Series hardware, accelerated with Intel Extension for PyTorch. Intel oneAPI Base Toolkit is used to optimize workloads across heterogeneous compute, and Intel Tiber AI Cloud supplies scalable cloud resources so the pipeline can run on systems that would not normally support models of this size. Generated sphere maps are inverted into BRDF-based materials, then used to extract or directly generate textures for downstream scientific rendering workflows.
Text prompts, sphere map examples, lighting-direction control images, and rendered training grids are fed into a Stable Diffusion plus custom ControlNet pipeline running on Intel GPUs. The model generates sphere maps under controlled lighting, which are then inverted into BRDF-based materials; textures are either extracted from the sphere maps through re-synthesis or generated directly, and the resulting materials are exported for scientific visualization workflows.
"Compared to CPU inference on a top-of-the-line Intel® Xeon Platinum processor (18.98s, 1.36 iterations/s) this constitutes 6x the performance"
"Our approach combines Stable Diffusion with a custom-trained ControlNet model, leveraging Intel’s compute hardware in conjunction with the Intel Extension for PyTorch for training and inference"
"The Intel® oneAPI Base Toolkit allows seamless optimization of compute workloads, while Intel® Tiber™ AI Cloud provides the high-performance computing environment necessary to handle the complexity of our models."
"Using diffusion models and custom control networks, we create multiple sphere maps, each lit from distinct angles."
"Using the Intel® Data Center GPU Max Series, we can generate new material maps in as little as 3.18s (4.24 iterations/s)"
"we address a longstanding challenge in scientific visualization–the creation of convincing materials to apply to scientific datasets"
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Intel Sample Sep 3, 2024 University of California Berkeley UC Berkeley needed to make generative LLM inference more efficient through low-precision quantization, but its SqueezeLLM approach depended on custom CUDA kernels that took weeks to implement and optimize and would be… 1.296 seconds SqueezeLLM 4-bit (0% sparsity) latency Intel Data Center GPU Max Series · Intel Tiber Developer Cloud · Intel oneAPI Base Toolkit 
Microsoft Sample Oct 15, 2026 Bar-Ilan University Students who miss lectures struggle to reconnect new lessons with prior material when course information is spread across notes, recordings, slides, and readings, and lecturers need a reliable way to answer routine… 71 courses Azure AI Services · Microsoft Azure · Microsoft Copilot Studio 
Google Cloud Sample Modified · Oct 9, 2026 Mindvalley Mindvalley needed to scale personalized AI learning companions for a large active-user base while keeping latency, uptime, and compute costs under control across customer-facing conversational workloads. 1.3 million annual agent requests BigQuery · Cloud Run · Cloud SQL 