Michael Hofmann

- Multitask Perception for video-based driver assistance systems - Sub-Project Manager for Neural Architecture Search and the Bosch Learning Platform - Agile Master Semantic Segmentation

Semantic Segmentation for Free Space Detection

Physically-based Rendering with MDL (Material Definition Language) Realtime Rendering Effects (e.g. volumetric water, localized reflection maps, ...) Rendering Performance Optimization (e.g. Occlusion Culling) for Massive Data on HMD Renderer Architecture (e.g. data-driven Rendergraph, dynamically-loaded Render Backends)

Machine Learning (Traffic Sign Recognition, Geometry Classification) Rendering (Raytraced Real-time radar simulation) Computer Vision (Master thesis - Advanced Variational Methods in Dense Monocular SLAM)

Extension of the PBRT-Renderer to enable detailed analysis of different integrators, integration of parametrized BRDFs into the Mitsuba-Renderer for visualizing internal light stage data sets using global illumination algorithms, implementation of a Fourier descriptor-based noise reduction algorithm, implementation of a CUDA-based discrete ordinate method (DOM) solver for volumetric light transport simulation

Integaration of the Oculus DK2 into research prototype renderer, Maya plugin for scene conversion and integration of RTT render back-ends, Further development of a projective augmented reality plugin

Implementation and evaluation of high quality texture filtering algorithms, Oculus Rift DK1 Integration in RTT DeltaGen, Development of RTT Stellar marerials for the integration into Bunkspeed Shot, Development of a projective Augmented-Reality Plugin for RTT DeltaGen in cooperation with Extend3D in the context of the ARVIDA project, Implementation and evaluation of different real-time subsurface scattering algorithms regarding physical accuracy for a Bachelor thesis

Computer Vision, Machine Learning, Rendering

(Mobile) Game Development, DirectX Programming, Procedural Modelling

Computer Graphics, Image Processing, Softwareentwicklung