Studentische_r Mitarbeiter_in AIT AENEAS Project

Semi-procedural generation of guard rail structures in Blender for Machine Learning data generation

The AIT Business Units “Transportation Infrastructure Technologies” and “Assisted and Autonomous Systems” offer an open Internship position (02/2021 – 05/2021) within the AENEAS project, where identification and classification of Vehicle Restraint Systems (VRS) are targeted. Due to recent availability of open-source scriptable photorealistic rendering solutions (such as Blender), new opportunities have arisen to artificially generate object-specific image data.
Internship activity
During the internship the candidate shall develop a scripted guard rail generation and rendering pipeline from a rich set of geometric parameters. The development shall rely on the Blender Python interface and it shall allow for the generation, rendering and compositing of semi-synthetic image datasets, subsequently used for learning representations of VRS parts. Ultimately, via these learned representations, characteristic guard rail parts such as the beam wave structure, base/post elements shall be recognized with a high accuracy from real images.

Expected Results

Based on our preliminary Blender data generation scripts (handling cameras, textures and illumination variations), the successful candidate shall elaborate a fully parameter-driven rail-guard generation, rendering and compositing pipeline (as a set of Python scripts). During development the candidate will also closely interact with the team performing the corresponding Machine Learning tasks.

Expected qualifications and interests

- Knowledge of Python, basic (preferably advanced) Blender concepts,

- Interested in turning exciting ideas into real Machine Learning based image analysis solutions in a challenging context


Csaba Beleznai, Center for Vision, Automation & Control, Assisted and Autonomous Systems,