Status: In Progress
- November 30, -0001
- - November 30, -0001
3DMotaaS aims to advance proof-of-concept technology for 3D motion search and retrieval through a strategic collaboration between Algolysis Ltd (SME/Host) and the CYENS – Centre of Excellence (RO/Partner). Leveraging the two Partners’ complementary expertise, know-how, and Intellectual Property, the consortium aims to develop and validate a 3D motion data search engine, which will in turn power a novel state-of-the-art Motion-as-a-Service prototype (TRL5).
Algolysis Ltd brings into the project its proof-of-concept motion search and retrieval technology (TRL3) and know-how, while CYENS invests in applying its expertise in Machine Learning and makes available its significant Motion Capture infrastructure necessary for the success of this endeavor.
One of the key features of this novel search engine is the ability to query-by-example, i.e., to search within a corpus of motion data using a motion sample, instead of keywords. To unlock the potential of this technology and benefit a multitude of sectors, the motion similarity engine can be used programmatically by third-party developers/applications, making it possible to embed motion search and retrieval in software targeted for entertainment (game development, film production), sports, health, robotics, biomechanics, etc.
Motivation and General Objectives:
Motion capture (aka mocap) is the standard method for digitizing human movements, which can in turn be used to animate 3D characters. It offers an alternative to the painstaking process of interactive computer animation for 3D characters. Although it has in the past been an exclusive ability of high-spending motion capture studios and labs, it is becoming more affordable and widely accessible. Marker-based systems that track physical markers on the actors are slowly giving way to markerless systems that use software to track an actor’s motion. It is nowadays possible to extract 3D motion with a snap of a button using a commodity smartphone. Quality, of course, is variable, and there is ample space for improvement, but the future of motion capture is markerless, and with the emergence of AI and quantum computing, that is becoming increasingly possible. The ease of generating motion data in conjunction with the ever-increasing need for 3D motion data to power applications in entertainment, sports, health, robotics, etc. is set to result in a massive trove of motion data that may remain untapped without the necessary tools to process, analyze, search, and retrieve them.
Similar to how photos and videos have taken center stage in modern society during the past two decades, technologies incorporating animated 3D characters are set to shape our future. Thus, digital 3D motions and tools to store, index, search, retrieve and manipulate them may add fuel to and accelerate the emerging Extended Reality (XR) revolution.
In this project, ALGOLYSIS (SME/Host Organization) and CYENS (Research Organisation/Partner) will devise a globally unique innovation: a 3D motion data search engine. This disruptive technology has immense potential for application, similar to how text, image, and video search engines have transformed just about every economic, societal, and cultural aspect of our society. The state-of-the-art in motion capture and computer animation offers a multitude of solutions for individual motion analysis challenges (e.g., motion retargeting, style analysis, action recognition, motion retrieval, etc.). However, a search engine for 3D motions remains elusive, mainly because it requires (a) combining know-how from different areas in ICT, and (b) development and integration of various motion analysis algorithmic components to address the entirety of issues related to a search engine’s implementation. The Consortium is strongly committed and highly motivated to advance the proposed search engine technology to TRL5 during this project.
Investigators
Efstathios Stavrakis Phd
(Algolysis Ltd)
Someone from Cyens?
(Cyens)
Partners
CYENS and Algolysis Ltd
Funding
