Andrei Baraian

I am a Ph.D. candidate in Deep Learning at the Image Processing and Pattern Recognition Group Lab, Technical University of Cluj-Napoca, under the supervision of Prof. Sergiu Nedevschi. My research focuses on Extending the generalization capability of deep neural networks.

I earned my Bachelor’s degree in Computer Engineering from TUCN in 2019, where I worked on improving a Structure from Motion pipeline by integrating semantic constraints, specifically for drone navigation.

In the fall of 2019 I have moved to Oulu, Finland, to pursue a Master’s degree in Artificial Intelligence at University of Oulu. At the same time, I was working at VTT Technical Research Center of Finland as a research trainee. My Master’s thesis explored the feasibility of utilizing Deep Neural Networks for estimating particles size distribution in mineral slurry. The work was part of the jointly funded project APASSI.

After completing my degree, I continued at VTT as a Research Scientist, focusing on Machine Vision for industrial applications across agriculture, plastics, defense and medical imaging. A notable contribution during this time was my involvement in the AHMED project, where we proposed a MLOps framework for developing AI/ML solutions for medical devices. Feel free to check AHMED report.

In June 2023, I embarked on a motorcycle journey across South America on my motorcycle, fullfiling one of my life-long dreams. A blog about those adventures will be soon out.

Since 2023, I have also been involved in BrightR, a startup developing digital dental shade matching technology.

Selected publications

Automatic Fracture Detection and Characterization in Borehole Images Using Deep Learning-Based Semantic Segmentation

Andrei Baraian, Vili Kellokumpu, Räty Tomi, and 1 more author

In Proceedings of the 18th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2023), 2023

18th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, VISIGRAPP 2023 ; Conference date: 19-02-2023 Through 21-02-2023

Abs

Fracture analysis represents one of the key investigations that needs to be carried in borehole logs. Identifying fractures, as well as other similar features (like breakouts or foliations) is essential for characterizing the reservoir where the drilling took place. However, identifying and characterizing the fractures from borehole images is a very time and resource consuming task, that require extensive knowledge from geological experts. For this reason, developing semi-automated or automated tools would facilitate and increase the productivity of fracture analysis, since even for one reservoir, experts need to analyze and interpret hundreds of meters of borehole images. This paper presents a deep learning based approach for application of automatic fracture detection and characterization in borehole images, relying on state-of-the-art convolutional neural network for accurate semantic segmentation of fractures. Target images consists of color borehole images, as opposed to acoustic or drill-core images, and uses real world data, both for training the deep learning model and testing the whole system. The system is evaluated by using multiple metrics and the final outputs of the system are the parameters of the sinusoids that define the predicted fractures.
Computing Particle Size Distribution of Mineral Rocks using Deep Learning-based Instance Segmentation

Andrei Baraian, Vili Kellokumpu, Janne Paaso, and 2 more authors

In 2022 10th European Workshop on Visual Information Processing, EUVIP 2022, Oct 2022
Improved 3D Perception based on Color Monocular Camera for MAV exploiting Image Semantic Segmentation

Andrei Baraian, and Sergiu Nedevschi

In 2019 IEEE 15th International Conference on Intelligent Computer Communication and Processing (ICCP), Oct 2019
Agile and Holistic Medical Software Development: Final report of AHMED project

Pasi Ahola, Andrei Baraian, Klaus Förger, and 10 more authors

Feb 2023

Abs

AHMED is a co-innovation project (https://www.regops.fi/) with six companies (Atostek, Bittium, Mylab, Solita, Taipuva Consulting and Terveystalo) and two research organisations (University of Helsinki and VTT). This report provides summarized results of the AHMED project focused on research, development and adoption of solutions for efficient production of data-driven applications for healthcare. The project particularly addressed challenges related to the usage of artificial intelligence, machine learning and personal data in regulated medical software. We present experiences and best practices for transition to the new medical device regulation (MDR), AI/ML compliant RegOps models, tools and toolchains for medical software development, healthcare register data usage for innovative applications and traceability approaches for research and AI/ML model development.