Karla Brkić, Ph.D.

▪ A custom solution for semantic segmentation in AR and VR applications, based on the DeepLab v3+ architecture.
▪ Prior to development, performed a full state-of-the-art review; designed, planned, and implemented an experimental framework to fully evaluate the client's needs and propose solutions while considering performance trade-offs.
▪ Implemented the semantic segmentation model for custom objects using TensorFlow and performed data preparation, augmentation, training, and evaluation.
▪ Performed model optimizations to achieve real-time performance and ported the model into Unity, successfully resolving performance challenges in the process.

▪ Researched, designed, planned, and implemented a deep neural network for volumetric 3D object reconstruction from 2D images.
▪ Integrated the network into a custom solution for rendered 3D object placement in real scenes using optimization techniques.

▪ Delivered detailed performance monitoring systems for crypto trading algorithms that provided various insights/visualizations.
▪ Spearheaded large-scale data analysis (price/volume data for hundreds of cryptocurrencies on several exchanges) and instituted new standard operating procedures for increasing workflow.
▪ Applied deep learning in crypto trading.

▪ Researched methods for protecting the privacy of individuals in surveillance videos by concealing or removing soft biometric and non-biometric identifiers while preserving data utility and/or naturalness.
▪ Integrated algorithms for pedestrian detection, segmentation, and frame-by-frame appearance alteration through the use of generative adversarial networks
▪ Enabled steganographic embedding of the original data in the de-identified video.

This was primarily built-in Python with components in C++.

▪ Used machine learning to develop a tool that aids the prediction of congestive heart failure.
▪ The tool analyzes short-term heart rate variability segments using hybrid feature selection and rotation forests.

▪ A solution for scanning multiple-choice exam answer sheets and automated grading using image processing techniques.
▪ Support for detecting annotated answer boxes as well as detecting and recognizing handwritten answer letters.

▪ Developed an automated traffic infrastructure inventory using computer vision techniques.
▪ Ideated and implemented advanced methods for reliably detecting and recognizing traffic signs in videos.
▪ Integrated the developed components with an existing geographic information system.
▪ Created system components for data acquisition and automated data annotation.

▪ Built a tool for automated tree ring detection, segmentation, and statistical analysis using computer vision and image processing techniques.
▪ Enabled performing high-precision measurements on images.

▪ Envisioned, designed, researched, and implemented a novel method for representing and understanding Spatio-temporal data (video data) in computer vision.
▪ Successfully worked with an industrial partner to tailor the developed method to traffic sign recognition in a commercial traffic infrastructure inventory, significantly outperforming other available solutions in terms of temporal accuracy.
▪ Ideated, wrote and published ten scientific papers at prestigious international conferences.
▪ Worked and thrived in a multicultural environment and maintained strong collaboration with the Graz University of Technology.

I co-authored over 40 scientific papers published in top-tier peer-reviewed conference proceedings and journals related to computer vision, machine learning, and artificial intelligence (IEEE CVPR workshops / IEEE conferences / Springer, Elsevier journals). See Google Scholar for a full list.

I served as a reviewer for top-tier scientific conferences and journals, including, e.g.:
▪ IEEE Transactions on Information Forensics and Security
▪ IEEE Transactions on Biometrics, Behavior, and Identity Science
▪ IEEE Transactions on Intelligent Transportation Systems
▪ IEEE Journal of Biomedical and Health Informatics
▪ IEEE International Conference on Robotics and Automation (ICRA)
▪ IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)
▪ IEEE International Conference on Intelligent Transportation Systems (ITSC)
• IEEE CVPR CV-COPS Workshop on the Bright and Dark Sides of Computer Vision
The full list is available upon request.