Research

I am a doctor in computer science working as a researcher at the West University of Timișoara.

On this page you can find a list of my research interests and my published papers. Also, you can find a link to my Ph.D. thesis and tools that I have developed.

Research Interests

As a computer science professor and researcher, I am passionate about artificial intelligence and its related fields, including machine learning, neural networks, and image processing. My extensive research in these areas has led to the publication of several papers and my Ph.D. thesis, for which you can find abstracts below.

In addition to my past research, I am currently working on an exciting project involving Tesseract OCR technology to transliterate old Romanian documents, called ROTLA. This project aims to develop a highly accurate OCR model and automate the process of transliterating national heritage documents.

More on ROTLA

If you're interested in learning more about my research or collaborating with me on a project, I'd love to hear from you. Feel free to get in touch with me.

Go to the contact form

PhD Thesis

In March 2022, I achieved a significant milestone in my academic career by successfully defending my Ph.D. thesis. My thesis focused on designing a nature-inspired simulator to model wind patterns and forecast cloud movements.

Over the course of many months, I spent countless hours poring over the literature, researching different methodologies, developing models, and testing and comparing results. This was a challenging but highly rewarding experience that has expanded my knowledge and expertise in the field of computer science.

I am proud to have accomplished this significant feat and look forward to continuing to contribute to the field of computer science through my research and teaching endeavors.

West University of Timișoara
Department of Computer Science

Cloud Movement Forecasting based on Wind Modeled from Satellite Imagery and a Modified Flocking Algorithm

  • Marius E. Penteliuc
  • Faculty of Mathematics and Computer Science
  • Department of Computer Science
  • West University of Timișoara, Romania
  • marius.penteliuc@e-uvt.ro

Abstract — Photovoltaic (PV) panels are increasingly used for generating power by transforming Sun irradiance into electric output. Transient clouds are absorbing irradiance and passing over a solar farm greatly affects its power production capability. PVs operate at their maximum capacity only when exposed to clear skies. Because of this PV output is variable and PV plants are hard to integrate into the power grid. Identification and forecasting of cloud movement are crucial for operators to take appropriate actions to meet energy demand such as starting backup generators and compensating with power reserves. Many methods rely on ground cameras and sensors, or satellite imagery, but they lack granularity, generalization, and assume idealistic linear motion of clouds over long distances. In the thesis, I propose and describe a solution to forecasting short-term cloud movement using only data obtained from satellite imagery without the need for gathering auxiliary information from other sources. It is based on a nature-inspired flocking behavior simulator and produces granular forecasts of cloud positions for the entire scene. It also generates a wind map that covers the scene and is continuously updated with each forecast. The algorithm is tested for scalability on multi-core CPUs and GPU cards, and performance is compared to a feed-forward back propagation neural network trained to forecast cloud movement from satellite imagery. The proposed solution provides short-term wind and cloud forecasts of up to several hours while being lightweight and using fewer resources than a compared neural network. The scalability of the implementation is tested and the speed-up values are up to 1000 times. Future development into a web service and the possibility to utilize images from all-sky cameras is discussed.

Tools

One of my greatest passions as a computer science researcher is building custom tools to aid me and my colleagues in our work. Over time, I have developed several such tools that have proven invaluable in my daily tasks.

The tools below are open-source and available on GitHub for anyone to use. I invite you to check them out and see how they can help streamline your work processes. Additionally, I welcome contributions from anyone interested in improving the functionality and usability of these tools.

Together, we can continue to push the boundaries of what is possible in the field of computer science and make meaningful contributions to the wider research community.

Transliteration Playground

A playground to test the OCR recognition model that we built for the ROTLA project, and it's transliteration ability. It features a simple interface that allows you to upload a PNG or JPG image, mark regions, recognize them using our model, and copy the transliterated text.

Try the web app View the source

Box Editor for Tesseract

A free web-based app that lets you edit Tesseract OCR LSTM box files in WordStr format. The app features automatic text line detection, tools for adding, editing, and deleting bounding boxes, and alphabet-based character highlighting. Additionally, the app automatically sorts the bounding boxes column- and row-wise for uploaded and generated box files.

Try the web app View the source

Transliteration Words Pairing

A free web-based app that lets you pair words from two different languages. It features a simple interface that automatically retrieves whole words from a box file and lets you pair them with words from another language. It allows downloading a word list and a dictionary with the paired words in JSON format.

Try the web app View the source
 

Published Papers

Throughout my career as a computer science researcher, I have had the privilege of publishing several papers in esteemed academic journals and conference proceedings. These papers cover a range of topics, including artificial intelligence, machine learning, and image processing, among others.

I am proud to have contributed to the advancement of knowledge in these fields through my research, and I invite you to explore the abstracts of my published papers. Each paper represents a unique perspective and a significant contribution to the broader conversation around cutting-edge research and innovation.

Whether you're a fellow researcher, a student, or simply interested in learning more about the latest developments in computer science, I hope you find these papers informative and engaging.

2023 IEEE International Conference on Big Data (IEEE Big Data)

Comparing ML OCR Engines on Texts from 19th Century Written in the Romanian Transitional Script

  • Marc Frîncu, Marius E. Penteliuc, Simina Frîncu, Gheorhe Bran, and Manuela Zanescu
  • Department of Computer Science
  • West University of Timișoara, Romania
  • {marius.penteliuc, ioana.frincu, gheorghe.bran, manuela.zanescu}@e-uvt.ro
  • †School of Science and Technology
  • Nottingham Trent University, United Kingdom
  • marc.frincu@ntu.ac.uk
  • ‡Central University Library Timișoara, Romania

Abstract — Many 19th century Romanian texts are written in the Transitional Script (RTS) combining Cyrillic and Latin characters in variable proportions depending on specific factors (period, region, publishing house, literary trends, and personal beliefs). Thousands of heritage documents in numerous libraries across Romania have yet to be digitized. Reading texts written in RTS is challenging for modern scholars, today’s students and OCR engines which due to specific particularities have difficulties in processing them. Transfer learning can be used to train state-of-the-art tools such as Tesseract or Transkribus, but their efficiency varies according to the underlying deep neural network, the training parameters, and the particularities of the scanned images and used scripts. In this paper, we compare the performance of two architectures on 180 OCR models for 5 scenarios (best CER <4.6% for Tesseract) and discuss OCR challenges on RTS texts.

Keywords — training; optical character recognition; transfer learning; training data; phonetics; data models.

DOI10.1109/BigData59044.2023.10386851

2023 Proceedings of the 4th Conference on Language, Data and Knowledge (LDK)

Challenges and Solutions in Transliterating 19th Century Romanian Texts from the Transitional to the Latin Script

  • Marc Frîncu, Simina Frîncu, and Marius E. Penteliuc
  • †School of Science and Technology
  • Nottingham Trent University, United Kingdom
  • marc.frincu@ntu.ac.uk
  • ‡Faculty of Mathematics and Computer Science
  • Department of Computer Science
  • West University of Timișoara, Romania
  • {ioana.frincu, marius.penteliuc}@e-uvt.ro

Abstract — During the 19th century, the Romanian script has undergone a massive yet uneven transition from the Cyrillic to the current Latin alphabet. The amount of existing literature written in that script as well as the problems it poses for OCR and transliteration engines make the problem highly challenging from a Big Data perspective. In this paper, we discuss the issues and propose and test a machine-learning solution trained on small datasets using either transfer learning from Latin/Cyrillic or from scratch.

MDPI Electronics, Volume 11, Issue 17 (September-1 2022)

A Solar Radiation Forecast Platform Spanning over the Edge-Cloud Continuum

  • Marc Frîncu, Marius E. Penteliuc, and Adrian Spătaru
  • †School of Science and Technology
  • Nottingham Trent University, United Kingdom
  • marc.frincu@ntu.ac.uk
  • ‡Faculty of Mathematics and Computer Science
  • Department of Computer Science
  • West University of Timișoara, Romania
  • {marius.penteliuc, adrian.spataru}@e-uvt.ro

Abstract — The prediction of PV output represents an important task for PV farm operators as it enables them to forecast the energy they will produce and sell on the energy market. Existing approaches rely on a combination of satellite/all-sky images and numerical methods which for high spatial resolutions require considerable processing time and resources. In this paper, we propose a hybrid egde–cloud platform that leverages the performance of edge devices to perform time-critical computations locally, while delegating the rest to the remote cloud infrastructure. The proposed platform relies on novel metaheuristics algorithms for cloud dynamics detection and proposes to forecast irradiance by analyzing pixel values taken with various filters/bands. The results demonstrate the scalability improvement when using GPU-enabled devices and the potential of using pixel information instead of cloud types to infer irradiance.

Keywords — solar irradiance; cloud dynamics; cloud type; metaheuristics; cloud–edge platform; smart grid.

DOI10.3390/electronics11172756

Full Text 21 MB
2021 IEEE International Conference on Big Data (IEEE Big Data)

Processing Large Satellite Imagery to Estimate Solar Irradiance

  • Marius E. Penteliuc
  • Faculty of Mathematics and Computer Science
  • Department of Computer Science
  • West University of Timișoara, Romania
  • marius.penteliuc@e-uvt.ro

Abstract — We use satellite imagery to identify cloud types and correlate with irradiance data from a solar platform. Processing the large images takes several minutes. Integrating into a web service would enable multiple users to estimate solar irradiance using just satellite imagery. It would also allow for the processing of many images at once as more remote sensing satellites are launched and data is easily accessible. The Sentinel 2 Cirrus Band shows an 81% correlation with the diffuse solar irradiance values measured at the platform.

Keywords — satellite imagery; irradiance; cloud types.

DOI10.1109/BigData52589.2021.9671916

2021 IEEE PES Innovative Smart Grid Technologies Europe (IEEE ISGT Europe)

Short Term Cloud Motion Forecast based on Boid's Algorithm for use in PV Output Prediction

  • Marius E. Penteliuc* and Marc Frîncu
  • *Faculty of Mathematics and Computer Science
  • Department of Computer Science
  • West University of Timișoara, Romania
  • marius.penteliuc@e-uvt.ro
  • †School of Science and Technology
  • Nottingham Trent University, United Kingdom
  • marc.frincu@ntu.ac.uk

Abstract — Forecasting cloud motion and dynamics is crucial for many areas of study. Solar energy production depends on the cloud coverage over the area which impacts PV output, by clouds limiting the incoming solar irradiance. In this paper we propose an adaptation of a nature-inspired technique called the Flocking Behavior Algorithm (or Boids Algorithm) for the problem of cloud motion forecasting, for nowcasting and intra-day prediction windows. Without limiting the application of the algorithm we validate it on a sequence of satellite images and show its efficiency for short term forecasting.

Keywords — cloud motion forecast; nature inspired algorithm; Boids Algorithm.

DOI10.1109/ISGTEurope52324.2021.9640027

2021 20th International Symposium on Parallel and Distributed Computing (ISPDC)

Parallel Cloud Movement Forecasting based on a Modified Boids Flocking Algorithm

  • Adrian Spătaru*, Larisa Tranca*, Marius E. Penteliuc*, and Marc Frîncu
  • *Faculty of Mathematics and Computer Science
  • Department of Computer Science
  • West University of Timișoara, Romania
  • {adrian.spataru, larisa.tranca96, marius.penteliuc}@e-uvt.ro
  • †School of Science and Technology
  • Nottingham Trent University, United Kingdom
  • marc.frincu@ntu.ac.uk

Abstract — Nowcasting is vital for PV farms in order to match supply and demand on energy markets. The accuracy of existing forecasting methods relies on short term modelling of cloud movement and dynamics by combining satellite images with ground based allsky observations. Cloud movement is modelled based on complex nonlinear numerical models which on the short term are outperformed by image analysis techniques. In this research we propose a simplified nature inspired model for nowcasting which scales with the data. The model is based on forecasting wind direction extracted from motion vector fields by modelling wind using a modified Boids Flocking algorithm. We show that the model is accurate in predicting the cloud coverage up to several hours ahead and that it scales with the number of particles on shared memory systems suited for commodity machines available to PV farm operators.

Keywords — Boids flocking algorithm; Cloud movement; OpenMP; CUDA.

DOI10.1109/ISPDC52870.2021.9521639

2019 21st International Symposium on Symbolic and Numeric Algorithms for Scientific Computing (SYNASC)

Prediction of Cloud Movement from Satellite Images using Neural Networks

  • Marius E. Penteliuc and Marc Frîncu
  • West University of Timișoara
  • Faculty of Mathematics and Computer Science
  • Department of Computer Science
  • Timișoara, Romania
  • {marius.penteliuc, marc.frincu}@e-uvt.ro

Abstract — Predicting cloud movement and dynamics is an important aspect in several areas, including prediction of solar energy generation. Knowing where a cloud will be or how it evolves over a given geographical area can help energy providers to better estimate their production levels. In this paper we propose a novel approach to predicting cloud movement based on satellite imagery. It combines techniques of generating motion vectors from sequential images with neural networks. First, the images are masked to isolate cloud pixels, then Farneback’s version of the Optical Flow algorithm is used to detect motion from one image to the next and generate motion vector flow for each pair of images. After that, a feed forward back propagation neural network is trained with the vector data derived from the dataset imagery. Different parameters for the duration of the training, size of the input, and the neighborhood radius of one point in the scene are used. Promising results are presented and discussed to weight the potential of the proposed algorithm for forecasting cloud cover and cloud position in a scene.

Keywords — cloud motion; forecasting; neural networks; satellite imagery.

DOI10.1109/SYNASC49474.2019.00038