Computer graphics, a sub-discipline of computer science, is traditionally concerned with algorithms for digital synthesis and manipulation of visual and geometric content. However, in the modern context, it is a multidisciplinary field that includes real-time and photorealistic rendering, geometric modeling, and real-time simulation. It is closely related to computer vision, which analyzes visual data, and intersects with other domains like machine learning and human-computer interaction. The outcomes of graphics research are widely applied in virtual effects, computer animation, video games, visualization, virtual and augmented reality, or computer-aided design software used in architecture, engineering, and other industries.
The Computer Graphics Group at IDEAS focuses on advancing the frontiers at the intersection of graphics, vision, and machine learning. Collaborating with renowned academic institutions and industry partners and actively contributing to top-tier research in the ACM SIGGRAPH and EUROGRAPHICS community, we aim to push the boundaries by developing novel techniques and methodologies. Our research spans both fundamental and applied aspects of 3D rendering, geometric modeling, and animation and simulation, all in conjunction with artificial intelligence (AI) development. AI’s role across sciences and industries is rapidly growing and is a key technology of the future.
Neural rendering
In computer graphics, a primary objective is to create photo-realistic and controllable images and videos in a process called rendering. Over the years, graphics developed advanced methods for global illumination and handling complex materials, resulting in accurate simulations of light transport in a scene. Recently, research communities introduced the concept of neural rendering, which extends traditional computer graphics principles by using elements of computer vision and machine learning. Neural rendering pipelines learn to render and represent scenes from unordered or structured images or videos. This approach allows for high-quality, realistic image generation from new viewpoints. Our research aims to explore this technique for synthesizing high-quality visual content for applications like video production and the gaming industry.
Neural modeling
Another key aspect of computer graphics is dealing with 3D geometric content, which includes geometric modeling, geometry processing, and shape understanding. Designing 3D geometric models is a complex process that requires both artistic and technical skills. However, current design software often falls short of the demands, leading to the need for creative workarounds. Our goal is to research new algorithms and develop novel solutions for geometric content generation using generative machine learning methods. That approach allows the creation of geometry in combination with other aspects, like physical properties and semantic functionality, and scales also to large geometric assets, like open and procedural worlds. These solutions can be applied in entertainment industries like virtual production and computer games, but also in other fields, like computer-aided design, engineering, and manufacturing (CAD/CAE/CAM).
Neural simulation and animation
Our research extends to neural simulation and animation, where we blend physical simulation and animation principles with machine learning capabilities. This approach is becoming an essential tool for investigating complex real-world phenomena, such as climate change or the behavior of crowds of humans in densely populated areas. Physical simulations provide a safe virtual environment for testing various scenarios, aiding decision-making processes. We employ neural networks to transform these problems into optimization tasks, which can be more data-efficient. Our research in this area has applications in visual content production, computer games, urban planning, and serious simulations such as security and medical training.
In summary, the Computer Graphics Group at IDEAS NCBR is committed to advancing the field of computer graphics, with a particular focus on the intersection of graphics, vision, and machine learning. We are excited about the potential of our work to shape the future of society and industries.
Google Scholar
Adam Karczmarz