Youbiquo, together with BI-REX Competence Center and with the support of Cineca, submitted a project that has been selected as part of the 2nd Open Call FF4EuroHPC.
FF4EuroHPC is a research and development project co-financed by the European Commission under the H2020 program and supported by EuroHPC JU.
The key concept behind FF4EuroHPC is to demonstrate to SMEs how they can benefit from using advanced HPC services for state-of-the-art ICT solutions. Through two open calls, innovative HPC experiments were selected that address the business challenges of SMEs. The first wave of experiments are underway and they aim to solve a number of business challenges with the help of HPC, HPDA, AI or ML.
The 2nd Open Call ended recently and the evaluations of the proposals, aimed at a wide variety of industrial sectors across Europe, were communicated. The second wave of experiments will begin in early 2022. The results will be presented through success stories, inspiring SMEs in Europe for further HPC-based innovation.
High Performance Computing (HPC) is the ability to process data and perform complex calculations at high speed. For comparison, a laptop or desktop with a 3 GHz processor can perform around 3 billion calculations per second. While this computation speed is considerably higher than what a human being can achieve, it is nothing compared to that of HPC solutions that can perform millions of billions of computations per second.
One of the best known types of HPC solutions is the supercomputer. A supercomputer contains thousands of compute nodes that work together to complete one or more tasks. This process is called parallel processing. It’s like having thousands of networked PCs combining computing power to get things done faster.
HandyTrack – HPC for Hand Gesture Dataset Generation and Deep Learning Training for Detection & Tracking is the Youbiquo project winner of the 2nd Open Call FF4EuroHPC dedicated to Artificial Intelligence and specifically to Computer Vision and Machine Learning applied to the use of HPC for training neural networks and for generating data based on 3D models in order to reduce the time required for training algorithms and allow for more accurate performance.
The experiment aims to generate a large video data set for hand gestures using realistic 3D hand models as well as behavior trees to generate gesture variations. Hence, it aims to train machine learning algorithms for hand gesture recognition. HPC will allow this process to take much less time than classical methods.
In Augmented Reality, it is necessary to manage virtual 3D objects in space. In real life we use our hands which have 6 DoF (3-axis rotation and 3-axis movement) and fingers for fine movements. If there was a natural way of control, it should be maintained and not replaced by a control device.
The transition from touchscreens to touchless interfaces and from real world entities to multiple mixed and virtual entities creates new meaning for our hands. The need for touchless interaction creates an opportunity to take advantage of hand movements and gestures. The worlds of mixed and virtual realities require the replacement of traditional input modes, which allow us to manipulate virtual objects in a natural way. The challenge is how to address the opportunity of hand tracking on existing computational devices with low power consumption and zero latency.
The goal of this HPC experiment is to design and develop a software library based on Machine Learning techniques to perform dynamic detection and recognition of hand gestures without interruption.