Publications
Bachelor Thesis
- Problem (dataset shift): training with purely synthetic data (computer rendered images) significantly reduces the object detection rate in real world industrial scenarios (initial tests obtained a mean average precision lower than 50%).
- Contribution: following the domain randomization approach, the performance of neural networks during the detection task achieved an mAP slightly over 60% on real world scenarios.
April 2023 HTW Saar (Saarbrücken, Germany) bachelor thesis written within Evocortex GmbH robotics company under the advisory of M.Eng. Waldemar Haag in Nürnberg, Germany.
Conference Paper
2022 IEEE Global Engineering Education Conference (EDUCON)
Journal Paper
2024 Educational Sciences volume 14
Targeted for IROS 2025:
Synthetic data generation for object detection and image segmentation tasks within underwater and hazing environments Working under the advice of Dr. Alberto Muñoz.