Bachelor Thesis

2023

In my bachelorthesis I created a pipeline for the easy creation of simulated hand-object interactions to generate a full dataset of synthetic training data. The training data was used to train a Neural Network for object pose estimation from RGB Images.

A central part of my pipeline was a Pose Editor, that allowed me to load in preprocessed objects and dynamically generate hand meshes grabbing that object. The grabbing pose is generated by an inverse kinematics solver for two seperate parametric hand models.