Masterarbeit Vision-Centric 4D Occupancy World Model (all genders)
Core
Develop a vision-centric Predictive Occupancy World Model using Transformer architectures to forecast future semantic 4D occupancy from sequential multi-view camera inputs for dynamic robotics like autonomous driving.
Role type
Master's thesis researcher (Computer Vision / Robotics)
Builds
Scalable 4D Occupancy predictions and semantic feature alignment pipelines
Domain
Robotics, Autonomous Driving, 3D Computer Vision
Deliverable
production ML models
Required skills
Python, PyTorch, 3D Computer Vision, Semantic Segmentation, Occupancy Networks, Vision Transformers (ViT), Foundation Models (DINO, CLIP), Self-/Weakly-Supervised Learning
Preferred skills
3D Gaussian Splatting, Knowledge Distillation, Multi-View Learning
Technologies
PyTorch, DINOv2, CLIP, SAM, nuScenes
Responsibilities
Develop Transformer-based Occupancy World Models, build and train pipelines with semantic feature alignment, conduct benchmarks on large-scale datasets for prediction accuracy and label efficiency
Seniority
Master's student (Research) (via careerplan.io/jobs/2624486-masterarbeit-vision-centric-4d-occupancy-world-model-all-genders-at-xitaso-gmbh-it-softwar)
