Imagine you have a massive pile of medical scans or 3D point clouds from a building survey, and you need to spot patterns that aren't obvious — like a subtle tumor shape or a structural defect. Traditional methods look at data points one by one, but some patterns only show up when you look at how distant data points relate to each other. NoMADS built math and software that captures these long-range connections and runs them fast on graphics cards, turning what used to be slow academic experiments into tools that can handle real-world data sizes.
Companies processing large volumes of complex data — medical images, 3D point clouds, satellite imagery — often miss important patterns because standard tools only look at local relationships. Scaling advanced pattern-detection algorithms to real-world data sizes is computationally expensive and typically requires deep mathematical expertise that most engineering teams lack.
The project built an open-source GPU-accelerated library for graph-based data models with GPGPU parallelization support. They also developed mathematical theory and efficient algorithms for nonlocal data processing, with demonstrated applications in biomedical imaging and point cloud processing.
If you are a medical imaging company struggling with noisy scans or inconsistent segmentation results — this project developed an open-source GPU-accelerated library for graph-based data processing that improves image analysis quality and scales to large datasets. The consortium included 27 partners across 9 countries, with direct focus on biomedical imaging applications.
If you are a surveying or geospatial company dealing with massive unstructured point clouds from drones or laser scanners — this project developed resolution-independent algorithms that process discrete point clouds efficiently on GPGPUs, regardless of data size. Their open-source library handles the heavy lifting of turning raw 3D data into structured, usable information.
If you are a remote sensing company processing large volumes of satellite or aerial imagery for agriculture, forestry, or urban monitoring — this project built efficient computational techniques for nonlocal data analysis that scale well with input data size. With 7 industry partners in the consortium, the tools were designed with real-world deployment in mind.
The main deliverable is an open-source library for graph-based models on GPGPUs, meaning there are no licensing fees for the software itself. Your costs would be integration effort and GPU hardware. The project received EUR 1,111,500 in EU funding under MSCA-RISE, which primarily funded researcher exchanges rather than product development.
The project explicitly targeted scalability — their algorithms are designed to be 'resolution-independent' and 'scale well with the size of the input data,' with GPU parallelization support. However, this was a research-to-application bridging effort, so industrial stress-testing would need to be verified with the team.
The key software deliverable is described as 'comprehensive open source' with GPGPU parallelization support. This suggests permissive licensing, but the exact license terms should be confirmed with the coordinator at Friedrich-Alexander-Universität Erlangen-Nürnberg. Open-source status means you can evaluate and adapt the code freely.
The project aimed to bridge the gap between academic theory and practical application of nonlocal methods. They produced working software with GPU support, but the MSCA-RISE funding scheme focuses on staff exchanges and knowledge transfer rather than commercial product development. Expect research-grade code that needs engineering for production.
Based on the project objectives, these methods capture complex long-range relationships in data that traditional local methods miss — particularly useful for noisy biomedical images and unstructured point clouds. The mathematical foundations provide stability guarantees against noisy input data that purely heuristic approaches lack.
The project ended in August 2023. With 27 partners across 9 countries and 20 universities involved, there is likely an active research community around the tools. The project website at nonlocal-methods.eu may provide current status. Contact the coordinator for information on continued development.
The NoMADS consortium is large and internationally diverse — 27 partners from 9 countries including DE, ES, FR, IL, IT, NL, PT, UK, and US. However, it is heavily academic: 20 of 27 partners are universities, with only 7 industry partners (26% industry ratio) and just 3 SMEs. The coordinator is Friedrich-Alexander-Universität Erlangen-Nürnberg, a strong German research university. This composition is typical for MSCA-RISE projects, which prioritize knowledge exchange over commercialization. For a business looking to adopt these tools, the academic weight means strong theoretical foundations but likely limited production-readiness. The 7 industry partners suggest some real-world grounding, though their specific roles would need investigation.
Friedrich-Alexander-Universität Erlangen-Nürnberg (Germany) — search for NoMADS project lead in their mathematics or computer science department
Want to explore whether NoMADS graph-processing tools fit your data pipeline? SciTransfer can arrange a technical briefing with the research team and assess integration feasibility for your use case.
