Project R-15876

Parallax: Cost-effective, photorealistic 3D content creation using Artificial Intelligence (Research)

Virtual 3D experiences allow us to experience the world in a whole new way. However, creating high-quality 3D content remains expensive, time-consuming, and reserved for talented 3D artists. 360-degree cameras allow us to digitally capture the physical world, but such footage limits users' freedom of movement to looking around from the camera's point of view. These experiences lack the essential element of exploration needed to feel truly present in a particular place—whether it's games in which we play the leading role, sports or concert streams that put us in the middle of the action, movies in which we are the hero, or Virtual Reality experiences that transport us to virtual tours and exhibitions. Recent breakthroughs in novel-view synthesis (NVS), such as Multi-Plane Images, Neural Radiance Fields (NeRF), and 3D Gaussian Splatting, offer a promising alternative. These technologies make it possible to create photorealistic 3D representations based on standard photos or videos, significantly lowering the threshold for 3D digitization. However, the adoption of novel-view synthesis is limited by a complex workflow and high hardware costs to process the images, as well as a lack of knowledge on how to achieve the best possible results. The aim of this project is to provide SMEs in Flanders with the right tools to digitize existing environments or objects in a cost- and time-efficient manner in photorealistic 3D using everyday cameras. These photos are processed using novel-view synthesis—AI algorithms that learn what an environment looks like from any viewpoint based on a few photos or videos of it. By introducing the right infrastructure and comprehensive yet targeted information, we want to eliminate the barriers to adopting this promising new technology.

01 October 2025 - 30 September 2027