Embedded Files
Free Volume Estimation using Monocular Vision in Logistics
Fraunhofer-Institut für Materialfluss und Logistik (Fraunhofer IML) February 2024 - May 2025
Free Volume Estimation using Monocular Vision in Logistics
Fraunhofer-Institut für Materialfluss und Logistik (Fraunhofer IML) February 2024 - May 2025
As part of the master’s thesis completed at Fraunhofer in partnership with DB Schenker, this project developed a computer vision-based solution for estimating free volume inside shipping containers using monocular images from mobile devices. The goal was to optimize cargo loading and warehouse management by replacing costly manual measurements with a scalable, deep-learning-driven approach. The work focused on fine-tuning monocular depth estimation models on domain-specific data, reconstructing watertight meshes from single-image point clouds, and benchmarking the results against traditional methods. This research demonstrates the feasibility of monocular depth estimation in logistics, contributing to both computer vision and supply chain efficiency by enabling precise and cost-effective volume estimation.
Anomaly Detection in Time Series Data for Wire-Cutting Process
Institute for Production Systems (IPS), TU Dortmund April 2024 - June 2024
Anomaly Detection in Time Series Data for Wire-Cutting Process
Institute for Production Systems (IPS), TU Dortmund April 2024 - June 2024
A joint project between TU Dortmund's Institute for Production Systems in collaboration with Bosch focuses on enhancing predictive maintenance for wire-cutting machining systems. We detect and predict anomalies by analyzing spectrograms of high-frequency torque data collected during the wire-cutting process to train neural networks for classification. Advanced statistical methods like t-SNE and clustering algorithms aim to identify patterns indicating potential failures, enabling timely maintenance and ensuring optimal equipment performance. This project integrates data analysis with practical industrial applications to boost production system reliability.
Technology-driven Innovative Solutions using Generative AI for BvB
Chair of Digital Transformation, TU Dortmund October 2023 -February 2024
Technology-driven Innovative Solutions using Generative AI for BvB
Chair of Digital Transformation, TU Dortmund October 2023 -February 2024
This project, in collaboration between the faculty of digital transformation and BvB Dortmund's online merchandising team, involved applying design thinking and digital innovation methods to create value for BvB using generative AI.
A working prototype was built and presented to the company, taking into account the customer's needs, market potential, and technological feasibility. The prototype was built on Microsoft Azure using Language Studio. It consists of a chatbot that collects customer feedback, provides question-answering services, and helps customers choose the correct sizes from their online merchandising stores.
Multi-View SfM for Historic Artifacts
Computer Vision Lab, Sapienza University, Rome February 2022 - April 2022
Multi-View SfM for Historic Artifacts
Computer Vision Lab, Sapienza University, Rome February 2022 - April 2022
As part of a project at Sapienza University, I created an open-source Structure from Motion (SfM) pipeline for reconstructing Roman Artifacts from museums around Rome using mobile devices. This concept allows students and archaeologists worldwide to capture, access and examine historical artefacts from Rome in 3D.
Stereo Reconstruction of Martian Geology
Image Processing Group, TU Dortmund October 2021
Stereo Reconstruction of Martian Geology
Image Processing Group, TU Dortmund October 2021
Generating 3D mesh models of the Martian surface using the images from the engineering cameras of Perseverance (Mars2020) and Curiosity (MSL). The project involved extensive work in image processing and 3D reconstruction using various methods such as stereo reconstruction and multi-view SfM.
The reconstructed surfaces were fused with a DEM (Digital Elevation Model) and integrated into a VR experience.
Industrial Bin Picking
Fraunhofer IML, TU Dortmund March 2021
Industrial Bin Picking
Fraunhofer IML, TU Dortmund March 2021
As part of a collaborative project between TU Dortmund and Fraunhofer IML, funded by the ML2R initiative, I contributed to the development of a vision-based bin-picking system for industrial robotics. My work focused on designing and implementing the computer vision, object grasping, and point cloud processing pipelines using the detectron2 framework for autonomous bin-picking. The system leveraged RGB-D data from a Zivid camera to perform precise object recognition and localization. Additionally, I helped create the DoPose dataset by developing a pipeline for generating 3D object models using the Iterative Closest Point (ICP) algorithm.
Grasping mechanism for UAVs MBZIRC 2020
Multi-Robot Systems, Czech Technical University January 2019
Grasping mechanism for UAVs MBZIRC 2020
Multi-Robot Systems, Czech Technical University January 2019
As part of my bachelor's thesis, I designed and implemented an active-reliable grasping mechanism for Autonomous Unmanned Aerial Vehicles (UAVs) to build cooperative walls for the Mohamed Bin Zayed International Robotics Challenge 2020 (MBZIRC 2020).
The team placed first in the challenge, and a short paper was published in Springer's (LNCS) lecture notes in computer science.
Lidar Integration for NIFTi
Centre for Machine Perception, Czech Technical University July 2018
Lidar Integration for NIFTi
Centre for Machine Perception, Czech Technical University July 2018
Worked with the Vision for Robotics and Autonomous Systems (VRAS) group on the NIFTi EU project to integrate the Velodyne VLP-16 LIDAR with the UGV (Robot) for mapping and localization (SLAM).
The LIDAR was added to assist the UGV in performing SLAM and localizing accurately under low visibility /hazardous conditions for search and rescue operations.
Google Sites
Report abuse