Aral Hekimoglu

Aral Hekimoglu

Machine Learning Engineer in Autonomous Driving @ BMW Group and TU, Munich

Sitemap

A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.

Pages

Posts

Future Blog Post

This post will show up by default. To disable scheduling of future posts, edit config.yml and set future: false.

Blog Post number 4

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 3

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 2

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 1

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

education

industry

Research Intern

  • Researched image processing algorithms for registration and segmentation to get familiarized with medical image processing concepts and tools used for building projects.
  • Created a user-interface plugin for MITK using C++ for medical image processing algorithms designed for use of health-care professionals working on healthcare data in detection of deformations.
  • Deployed the applications in Docker containers to eliminate environment dependency problem.

Computer Vision Research Intern

  • Designed a new person re-identification model to improve performance of the model in-use by +16% by proposing a part-based generator network to generate more realistic looking pedestrian images.
  • Organized experiments in PyTorch to compare performance and image quality that resulted in reduction in complexity of the backbone network which doubled the speed at test time.
  • Participated in entire R&D cycle: coding the network, data loading, training in a Nvidia-Docker container, conversion to a deployable TensorRT Caffe model, integrating with surveillance system in C++.

Computer Vision and Machine Learning Engineer

  • Researched Active Learning methods for data-efficient model development. Researched method helped 2D Object Detection deep learning model achieve peak performance with 30% less training data.
  • Constructed an Active Learning based pipeline consisting of data selection, training dataset creation and model training that helped engineers spend less time on data preparation processes and more on model development.
  • Worked on training and implementation of multiple autonomous driving perception models including: 2D Object Detection, Lane Detection, Point Cloud Segmentation, Perception Fusion and Motion Prediction.
  • Deployed various neural network models using TensorRT and C++ 11 for faster inference running on NVIDIA DRIVE AGX and ROS.
  • Participated in design and decision making process of full perception architecture for a self-driving car.

Doctoral Researcher

  • Conducted research at BMW while also working towards a Doctoral Degree from Technical University Munich
  • Developed Active Learning methods for Autonomous Driving tasks to increase efficiency of data selection process
  • Collaborated with other PhD students in the team and supervise master thesis students

publications

Active Learning for Object Detection with Non-Redundant Informative Sampling


Efficient Active Learning Strategies for Monocular 3D Object Detection


Multi-Task Consistency for Active Learning


Monocular 3D Object Detection with LiDAR Guided Semi Supervised Active Learning


Active Learning with Task Consistency and Diversity in Multi-Task Networks