2024 mai võitjad
Bakalaureuse esimese aasta kategooria:
- Võrkpalli kotkasilm (WWW, PDF) - Philip Paškov, Oskar Männik.
- Programmeerimine II Hall of Fame (WWW, PDF) - Kevin Akkermann, Rainer Vana.
- Clown Breeding Game (WWW, PDF) - Elisabeth Serikova, Daria Savtšenko.
Bakalaureuseastme kategooria:
- Kiire Viibe: Estonian Fingerspelling Recogniser (WWW, PDF) - Hendrik Matvejev, Karl Kristjan Puusepp, Priit Peterson.
- I Know Your Password? - Website Paroolikratt (WWW, PDF) - Anett Pärismaa.
- KOIT (WWW, PDF) - Kevin Akkermann, Andry Avamägi, Carl Valgus.
Magistriastme kategooria:
- Stacks of Gold: Utilizing GANs to Enhance 3D Microscopy Imaging Data (WWW, PDF) - Dmytro Fedorenko.
- Computer Vision Meets Forestry: Multi-class Semantic Segmentation of Tree Species from UAV-Images (WWW, PDF) - Ali Zeynalli.
- Imagining Infinity: Endless CT Datasets through Conditional Diffusion Models (WWW, PDF) - Ekaterina Sedykh.
Parim Demo:
- Kiire Viibe: Estonian Fingerspelling Recogniser (WWW, PDF) - Hendrik Matvejev, Karl Kristjan Puusepp, Priit Peterson.
Publikupreemia:
2024 mai osalejad
Bakalaureuse esimese aasta projektid:
- B101 Programmeerimine II Hall of Fame - WWW, PDF
Programmeerimine II (LTAT.03.007) ülesannete edetabelite kokku võtmine ja visualiseerimine ajas.
Kevin Akkermann, Rainer Vana. - B102 Võrkpalli kotkasilm - WWW, PDF
Võrkpalli kotkasilm on projekt, mis tuvastab kasutaja soovitud võrkpalli mängu videol väljaspool väljakupiire maandunud palli kaadrid. Tarkvara kasutab masinõppel põhinevaid YOLOv5 ja YOLOv8 mudeleid koos OpenCV arvutinägemise ja masinõppe tarkvara teegiga.
Philip Paškov, Oskar Männik. - B103 Clown Breeding Game - WWW, PDF
Tere! Clown Breeding Game on mäng, mis on loodud Java ja JavaFX programmeerimiskeeltes. Mängu põhimõte on üsna lihtne: mängu alguses on mängijal ligipääs ühele maailmale, kus elab üks tegelane - kloun Peeter Paanika. Mängu alguses saab igaüks lühikese juhendi: uue taseme klouni avamiseks on vaja ristuda kahe eelmise taseme kloune. Ristumiseks mõeldud kloune saab poest osta valuuta eest - klounide pisarad. Nende saamiseks tuleb klouni lüüa. Kui mängija avastab ühes maailmas 6 erinevat tüüpi klouni, avaneb talle järgmine maailm. Kokku on mängus saadaval 3 maailma. Samuti on tal ligipääs juba avatud tegelaste galeriile ja avatud maailmade vahel liikumise võimele.
Elisabeth Serikova, Daria Savtšenko.
Bakalaureuse projektid:
- B01 Andmebaasid keeleteaduslikuks uurimistööks - WWW, PDF
Sildikujul keeleteadusliku materjali uurimist ja haldust toetavad Wordpressi veebilehed, mis funktsioneerivad andmebaasidena.
Mihkel Roomet. - B02 Thesis Writing Simulator - WWW, PDF
A choice-driven life simulation game focused on physical and emotional well-being.
Mihkel Roomet. - B03 Estonian open data portal data analytics dashboard - WWW, PDF
Open data is data that can be used and shared by everyone for both commercial and non-commercial purposes. The Estonian Open Data Portal provides an opportunity to access Estonian Open Data to be consumed or visualized. Currently, the visualization tool available on the portal is not consistent, often doesn't open, or is missing altogether from the options. Additionally, it is outdated, and the user interface and functionalities are based on outdated technologies. To address these issues, a graphical user interface was developed in this bachelor's thesis, enabling the visualization and analysis of open data.
Kris Porovarde. - B04 Veebipõhise Linuxi käsurea õppekeskkonna tootestamine - WWW, PDF
Varasema veebipõhise Linuxi käsurea õppekeskkonna prototüübi edasiarendus, täiustamine ning üleviimine pilvetehnoloogiale ehk tootestamine.
Taavi Eistre. - B05 Under Pressure - WWW, PDF
Under Pressure veebirakenduse idee on võimaldada tudengitel Tartu Ülikooli õppeainetega seotud stressitaset jälgida ning võrrelda seda keskmise stressitasemega. Seeläbi on võimalik igal tudengil vaadata õppeaine stressitaset ning saada aimu, kuidas neil läheb võrreldes üldise keskmisega ning mis on vastava õppeaine kõige koormuserikkamad perioodid.
Hjalmar Vaiküll, Stefan Ehin. - B06 Kiire Viibe: Estonian Fingerspelling Recogniser - WWW, PDF
A user-friendly website for learning Estonian fingerspelling. Features an interactive interface that uses your webcam and a pretrained model to accurately recognise fingerspelling signs in real-time.
Hendrik Matvejev, Karl Kristjan Puusepp, Priit Peterson. - B07 syppi? / typsi? - A rating platform for beverages - WWW, PDF
Find the best drinks in town and let your opinion be heard. Search, filters, and sorting help you narrow your results. typsi? differs from syppi? only in the inclusion of alcoholic drinks. Remember, please drink responsibly.
Oliver Jõgar, Matias Jürgenson, Joosep Hubel. - B08 I Know Your Password? - Website Paroolikratt - WWW, PDF
The aim of the Bachelor thesis was to investigate the password creation habits and patterns of Estonian-speaking users. The study provides recommendations for avoiding predictable passwords and strengthening security practices. Estonian user passwords leaked online were collected, and a survey mapped password creation habits. A website was developed to test if a password appears in a 50-million-entry attack dictionary. The full dictionary (50 GB) helps identify vulnerable passwords and patterns for security testing.
Anett Pärismaa. - B09 Steinnriki - WWW, PDF
2D vertikaalne linnaehituse mäng.
Karm Koduvere. - B10 KOIT - WWW, PDF
Interaktiivne programmeerimise õppeplatvorm koos Pythoni ja JavaScripti õppematerjalidega. Lisaks veel õpetajapaneel, mis lihtsustab klassiruumis/MOOC-i vormis programmeerimise õpetamist.
Kevin Akkermann, Andry Avamägi, Carl Valgus. - B11 Laine funktsiooni kokkulangemise algoritmi laiendus mängumootoris Godot - WWW, PDF
Laine funktsiooni kokkulangemise algoritmi laiendus, mis teeb algoritmi kasutamise lihtsaks ja intuitiivseks.
Markus Männil. - B12 ATI juturobot - WWW, PNG
Tehisintellektil põhinev juturobot, mis aitab kasulikku informatsiooni leida ülikooli ATI domeenidega.
Kaarel-Richard Kaarelson, Rannar Zirk. - B13 EleBall - WWW, PDF
EleBall is a casual arcade videogame, where you control a circus elephant trying to catch objects, whilst also uncovering the story behind the EleBall circus.
Timo Jairus, Olaf Daniel Seisler, Maria Anett Kaha. - B14 Stellar Manager - WWW, PDF
Transpordi ja majanduse simulatsiooni mäng kosmoses!
Henrik Tamm. - B15 Injecting Inputs Over Internet Connections - WWW, PDF
System to control SIGINT games with a controller that sends gamepad inputs to the game host machine over a WiFi connection.
Raiko Valo.
Magistri projektid:
- M01 UE5 Flair's Integration UI, Shaders, and Image Algorithms for NPR - WWW, PDF
This project showcased how integrating Flair's into UE5 opens up to new possibilities for NPR creative expression in digital art and in game development.
Alicia Sudlerd. - M02 Predicting the molecular mechanisms of genetic variants - WWW, PDF
This project presents an original dataset for training and evaluation of the genetic variant mode of action prediction models and a proof-of-concept MoA model, classifying GWAS variants into two classes: splicing QTLs and gene expression affected by chromatin accessibility QTLs.
Dzvenymyra-Marta Yarish. - M03 Eesti ilutoodete andmebaas - WWW, PDF
Oleme kaks andmeteaduse tudengit Tartu Ülikoolist, kellel on sügav huvi keemia vastu. Oma õpingute käigus avastasime, kui keeruline võib olla ilutoodete koostisosade mõistmine. Paljudel inimestel puuduvad teadmised ja aeg, et analüüsida kõiki koostisosi, mis nende lemmiktoodetes peituvad. Meile tundus see probleemina, mis vajab lahendust.
Heili Aavola, Annabel Hiiu. - M04 Formidable Fortress - WWW, PDF
Artillery Game with Dynamic Difficulty Adjustment.
Euna Islam. - M05 Weakly Supervised Segmentation in Medical Imaging: A Counterfactual Approach - WWW, PDF
Our innovative Counterfactual Inpainting (COIN) approach inspired by the work of Singla et al. accurately segments pathology regions in CT scans without reliance on the existing segmentation masks.
Dmytro Shvetsov. - M06 Stacks of Gold: Utilizing GANs to Enhance 3D Microscopy Imaging Data - WWW, PDF
Confocal microscopy, a pivotal tool in biomedical research, offers detailed 3D visualizations of living cells, providing insights into their spatial morphology, interactions, and life cycle progression. Fluorescence (FL) microscopy images are of high quality; however, they require expensive and toxic FL labeling to be captured. We focus on extracting detailed 3D information from easy-to-perform but low-quality bright-field (BF) images via BF-to-FL translation and enhancing the quality of 3D FL microscopy images through deconvolution, denoising, and deblurring.
Dmytro Fedorenko. - M07 Imagining Infinity: Endless CT Datasets through Conditional Diffusion Models - WWW, PDF
Medical imaging, the technique of creating visual representations of the interior of a body for clinical analysis and medical intervention, critically depends on the availability of extensive and high-quality datasets. However, the acquisition of such datasets is often limited by logistical, ethical, and privacy concerns. Diffusion models, known for their effectiveness in generating high-quality synthetic data, can mitigate these challenges by producing realistic medical images for model training and research purposes. This thesis addresses the challenge of data scarcity in medical imaging by exploring the efficacy of diffusion models in generating synthetic CT images that can be used to enhance dataset diversity and volume. Here, we demonstrate that diffusion models can effectively generate synthetic CT images that closely mimic real diagnostic images, thereby potentially expanding the breadth of available training data for medical AI applications. The results reveal that the synthetic images produced are not only almost indistinguishable from real images but also retain the necessary clinical details, which is an advancement over previous generative models that often sacrificed clinically relevant details. This work further exemplifies the utility of synthetic data generated by diffusion models in improving the training and performance of AI systems in diagnosing and analyzing medical images. The integration of diffusion models into medical imaging practices promises to significantly strengthen the AI-driven diagnostic tools. By providing a novel method for generating synthetic medical images, this research highlights the potential of advanced generative models in overcoming practical and ethical barriers in medical research.
Ekaterina Sedykh. - M08 Computer Vision Meets Forestry: Multi-class Semantic Segmentation of Tree Species from UAV-Images - WWW, PDF
Accurate tree species segmentation is a complex challenge in forestry with profound implications for sustainable forest management, biodiversity conservation, and understanding forest ecosystems. The goal of the project is to explore the potential of deep learning to automate the multi-class semantic segmentation of tree species from UAV images.
Ali Zeynalli.