Date and place: May 6th, 2:15pm at room T4 (Aalto CS) + zoom

Abstract: Advances in implicit 3D representations are changing    the way we think about 3D geometry and 3D computer vision. Advances in neural representations and differentiable rendering are changing the way we capture photo-realistic scenes, solve for unknown camera and scene parameters, and integrate new 3D representations into classical computer vision and computer graphics pipelines. These advances are introducing new methods for visual artists and the film industry to generate photo-realistic 3D scenes in the animation and gaming industries. This talk introduces the topic of NeRFs and 3D Gaussian Splatting representations in the context of inverse rendering with specific applications in problems within computer vision.

Bio: Matias Turkulainen is a MSc student interested in the intersection of computer graphics and computer vision. Matias is a soon-to-be-graduate in Robotics from ETH Zurich and is pursuing a PhD with Juho Kannala exploring new geometry representations for 3D reconstruction with applications in computer vision problems. He is an active open-source developer, mainly working on the collaborative Nerfstudio project, and is interested in topics related to implicit 3D representations and inverse rendering.