Research & Development
Cyanapse provides on-demand R&D expertise to solve complex technical challenges across AI, computer vision, data science, and compute infrastructure. Backed by a track record of innovation since 2016, we deliver our most consequential work at the intersection of deep technical expertise and principled methodology.
Cyanapse has been conducting and investing in applied R&D since 2016, with formal recognition from Innovate UK and a history of resolving genuine scientific and technological uncertainties in computer vision, AI and computing infrastructure. We offer this R&D expertise as a service, bringing the rigour, methodology and intellectual depth of an R&D function to clients who need to solve hard problems with no available out-of-the-box solutions.
Applied Innovation with a Track Record
We offer R&D (Research and Development) as a service to clients who are ready to invest in genuine innovation and move from uncertainty to capability, including AI integration, roadmaps, exploratory analysis, production-grade data and AI pipelines, high-performance cloud infrastructure, cutting-edge bespoke image processing systems, and more.
We have been pioneering R&D since 2016. Our work has been awarded a prestigious Innovate UK grant, presented at international conferences, and published in peer-reviewed scientific journals.
R&D is not always a straight line: It is the patient, rigorous and often humbling work of confronting genuine scientific and technological uncertainty and advancing the boundary of what is possible.
Our R&D work spans a wide spectrum ranging from data science to deep learning. This is illustrated with our own projects at the frontier of neuroscience and computer vision, where we have been building the compute infrastructure to make our algorithms run in real time, tuning the data pipelines that feed them, and iterating on the methodology until the results were worth publishing.
Our R&D Methodology
Our R&D methodology is grounded in the BEIS (2004) guidelines for science and technology. We identify genuine technological uncertainty, establish a rigorous baseline, pursue systematic resolution, and document progress and outcomes. Every R&D project we undertake begins with a feasibility assessment that distinguishes between problems that can be solved with engineering and problems that require research.
Case Study: Product Digitisation and Texture
Augmentation
In collaboration with a partner, we developed a fully automated four-step 3D reconstruction process for product photography, incorporating our photorealistic lighting correction and texture augmentation technology. The pipeline extracted, corrected and augmented textures from photographed consumer goods, enabling photorealistic digital twins of physical products at scale. This work represents a major technological advance in automated image editing.
Case Study: Applied Data Analysis
We have been engaged on projects for data exploratory analysis, data presentation and prototype development. Our work included data cleaning and preprocessing, statistical profiling, machine learning prototype development, and the production of data visualisations and analytical reports for use in specialised conferences. These projects exemplify our capacity to bring R&D-grade analytical rigour to real-world datasets with tangible deliverables, scientific dissemination and social impact.
Case Study: Photorealistic Style Transfer
In October 2018, Cyanapse presented at the NVIDIA Worldwide GTC EU Conference in Munich, Germany. Our presentation covered our advances in computing algorithms for real-time photorealistic lighting transformations and the benchmarking methodology we developed to demonstrate their performance. This was the first public presentation of our cloud computing infrastructure approach and its application to computer vision tasks at consumer latency.
Case Study: Perceptual Quality Assessment Metrics
We developed a proprietary perceptual quality assessment metric, the subject of a cited 2022 publication in cs.CV, that correlates directly with real human assessments of image quality. Built on a fusion of standard metrics including SSIM, PSNR and TV ratio, it serves both as an optimisation objective during model training and as a quality gate during inference, ensuring that only results meeting a defined perceptual threshold are delivered. The metric produces an absolute quality score for individual images, making it equally useful for automated image selection, model evaluation and editorial quality control.
R&D Recognition
Cyanapse was selected as a member of the Machine Intelligence Garage, a national programme funded by the UK government, in 2018. We were one of five UK-based companies invited to participate in the Digital Catapult 5G Test Bed in Brighton (UK) in 2018-2019, where we demonstrated our real-time computing infrastructure for image processing. Our R&D work was awarded Innovate UK Smart Grants in 2019-2021 and formally assessed and recognised as qualifying R&D under the BEIS (2004) guidelines by HMRC. We are members of the NVIDIA Inception Programme, supporting AI startups at the frontier of deep learning and GPU computing, and recipients of AWS Activate credits in support of our cloud infrastructure research. We collaborated with researchers at the University of Sussex, whose work in neuroscience and engineering intersects directly with our vision research programme. Our published research has been cited by MLJAR Research, a leading data science company behind the popular AutoML framework, and our initiatives have been recognised by the UK Government as part of their reported achievements in the AI Sector Deal.
Other R&D Projects
Visit the Cyanapse Vision Lab page to discover our current and previous research projects.