2023 mai võitjad
Bakalaureuse esimese aasta kategooria:
I - Genekas - TÜ tunniplaani generaator - WWW, PDF
Genekas on veebileht, mis koostab õpilase sisestatud tundidest kõikvõimalikud tunniplaanide kombinatsioonid, arvestades rühmade toimumisaegu ning kasutaja eelistusi.
Johan Kirikal, Daglas Aitsen.
II - Bibliokirjete mestimine - WWW, PDF
Projekti eesmärk on dubleeritud bibliograafiliste kirjete tuvastamine ja liitmine.
Martin Pajuste.
III - 1. kursuse AI jurist - WWW, PDF
AI jurist on AI-põhine tööriist, kes teeb oma parima, et Eesti seadustega seotud info oleks kõigile kergesti kättesaadav. Meie programm võtab CSV failina sisendiks Eesti Vabariigi põhiseaduse ning oskab selle põhjal vastata kõiksugustele juriidilistele küsimustele. Projekti saab ka väga lihtsalt kohandada töötama teiste CSV/PDF failidega.
Karl-Andreas Meus, Jan Erik Köst.
Bakalaureuseastme kategooria:
I - Plants disease - WWW, PDF
The project aims to classify the disease of most common crops as well as provides insights into the model decision-making.
Aleksandr Makarov, Nika Mikhailava.
II - Computer Vision Meets Microbiology : Deep Learning Algorithms for Classifying Cell Treatments in Microscopy Images - WWW, PDF
Cell classification is one of the most complex challenges in biomedical research that has significant importance to personalised medicine, cancer diagnostics and disease prevention. The aim of this research is to explore potential of deep learning to automate the classification of microscopy cell images into four cell treatments: Palbociclib, MLN8237, AZD1152 and CYC116.
Ali Zeynalli.
III - Baariabiline - WWW, PDF
Baariabiline on programm, mis on siiani olnud puudu igalt tudengipeolt. Tihti peale käib baaris arveldamine ainult sularahas. Sellega kaasneb aga tüütu tagastusraha. Programm lubab kliendil jätta baari krediiti, et selle eest õhtu jooksul jooke osta. Lisaks loeb programm kokku kõik õhtu jooksul ostetud joogid ja arvutab välja baari teenitud tulu. Link programmi lähtekoodile on veebilehel.
Annabel Hiiu, Lauri Lopp.
Eriauhind:
- B01 Mine Sweep Battle - WWW, PDF
A simple casual fast-paced online competitive game that is easy to pick up for anyone.
Mihkel Roomet, Kaarel Koido.
Magistriastme kategooria:
I - SwinIA: Self-Supervised Blind-Spot Image Denoising without Convolutions - WWW, PDF
The first model to prove the concept of a pure-transformer blind-spot image denoising network with only one forward pass, simple MSE loss, and no demand for a known noise model.
Pavel Chizhov.
II - "FiBar": a Tool for Automated Analysis of Complex Biomaterials from Microscopy Images - WWW, PDF
The goal of "FiBar" is to provide an automated microscopy image analysis tool for complex biomaterials. These complex biomaterials consist of fibers with encapsulated components, such as bacteria. The tool is divided into two pipelines - one for fiber diameter measuring, the other for bacterial cell analysis. The ultimate goal of this "FiBar" is to reduce the workload of the researcher by providing a viable alternative to the manual analysis.
Marilin Moor.
III - Veebirakendus programmeerimise õppimiseks II ja III kooliastmes - WWW, PDF
Programmeerimisõpik, mis on suunatud II ja III kooliastmele ning õpetab programmeerimist keeles Python.
Kristiina Keps.
2023 mai osalejad
Bakalaureuse esimese aasta projektid:
- B101 Bibliokirjete mestimine - WWW, PDF
Projekti eesmärk on dubleeritud bibliograafiliste kirjete tuvastamine ja liitmine.
Martin Pajuste. - B102 Blackjack - WWW, PDF
Blackjack kaardimäng JavaFX teegiga.
Oliver Jõgar, Romek Hanson. - B103 Andmebaas "Kunstiteosed" - WWW, PDF
Tegu on andmebaasiga, mis koondab erinevate kunstnike teosed, nende loomiseks kasutatud materjalid, kunstivoolud ja -žanrid ning praeguse asukoha muuseumi, linna ja riigi tasandil. Andmebaasi kasutaja saab mugavalt ja kiirelt näiteks kunstivoolu või kunstniku nime kaudu leida, kuhu tuleks reisida, et tuntud teoseid oma silmaga näha.
Iris Rand, Hedi Tamm. - B104 Delta Majaplaan - WWW, PDF
Interaktiivne majaplaan Delta õppehoone ruumide leidmiseks.
Oliver Pikani, Darius Kosk. - B105 1. kursuse AI jurist - WWW, PDF
AI jurist on AI-põhine tööriist, kes teeb oma parima, et Eesti seadustega seotud info oleks kõigile kergesti kättesaadav. Meie programm võtab CSV failina sisendiks Eesti Vabariigi põhiseaduse ning oskab selle põhjal vastata kõiksugustele juriidilistele küsimustele. Projekti saab ka väga lihtsalt kohandada töötama teiste CSV/PDF failidega.
Karl-Andreas Meus, Jan Erik Köst. - B106 Genekas - TÜ tunniplaani generaator - WWW, PDF
Genekas on veebileht, mis koostab õpilase sisestatud tundidest kõikvõimalikud tunniplaanide kombinatsioonid, arvestades rühmade toimumisaegu ning kasutaja eelistusi.
Johan Kirikal, Daglas Aitsen. - B107 Invest andmebaas - WWW, PDF
Andmebaas, mis võimaldab koondada kokku erinevad investeerimisvõimalused üle maailma, nende kohta käivad andmed läbi aegade ja aitab koostada investeerimisanalüüsi.
Hannabeth Hansen, Kelli Gagarina.
Bakalaureuse projektid:
- B01 Mine Sweep Battle - WWW, PDF
A simple casual fast-paced online competitive game that is easy to pick up for anyone.
Mihkel Roomet, Kaarel Koido. - B02 Baariabiline - WWW, PDF
Baariabiline on programm, mis on siiani olnud puudu igalt tudengipeolt. Tihti peale käib baaris arveldamine ainult sularahas. Sellega kaasneb aga tüütu tagastusraha. Programm lubab kliendil jätta baari krediiti, et selle eest õhtu jooksul jooke osta. Lisaks loeb programm kokku kõik õhtu jooksul ostetud joogid ja arvutab välja baari teenitud tulu. Link programmi lähtekoodile on veebilehel.
Annabel Hiiu, Lauri Lopp. - B03 Cross-Country Horse Run - Hyper-Casual Mobile Game - WWW, PDF
Cross-Country Horse Run is a 3D endless runner game developed for a Bachelor's thesis. The player has to run as far as possible with the horse and avoid obstacles in order not to lose lives and collect coins.
Käroly Raudsepp. - B04 Computer Vision Meets Microbiology : Deep Learning Algorithms for Classifying Cell Treatments in Microscopy Images - WWW, PDF
Cell classification is one of the most complex challenges in biomedical research that has significant importance to personalised medicine, cancer diagnostics and disease prevention. The aim of this research is to explore potential of deep learning to automate the classification of microscopy cell images into four cell treatments: Palbociclib, MLN8237, AZD1152 and CYC116.
Ali Zeynalli. - B05 Take a Leap: Hyper-Casual Mobile Game - WWW, PDF
Take a Leap is a hyper-casual mobile game that was made for a bachelor's thesis.
Grete Hollo. - B06 Progvaramu - WWW, PDF
Veebirakenduse abil on kasutajatel võimalik süsteemi lisada kursusega seotud dokumente. Rakendus võimaldab kasutajatel üles laaditud dokumente hallata ja filtreerida. Rakenduse sisselogimiseks vajaliku info leiab projekti repositooriumist, mis on lisatud postri juurde.
Kaur Taal. - B07 Artificial Thesis Detection - WWW, PDF
It is common these days to rely on assistance of chat bots like Chat-GPT. They are helpful and convenient tools to simplify everyday life. This project aims to classify Bachelor's works to be written by human or GPT-generated.
Danila Kuklianov, Alexander Makarov. - B08 Plants disease - WWW, PDF
The project aims to classify the disease of most common crops as well as provides insights into the model decision-making.
Aleksandr Makarov, Nika Mikhailava. - B09 Sihtleht mängule Blastronaut - WWW, PDF
Käesolevas bakalaureusetöös uuritakse turunduslike ja sihtlehele omaseid põhiprintsiipe. Selgitatakse nende tähtsust sihtlehe loomises. Tutvustatakse arvutimängu ostuprotsessi. Antud töös luuakse arvutimängule Blastronaut tulemuslik sihtleht järgides põhiprintsiipe, et suurendada mängu leitavust, inimeste teadlikkust ning mängu Steam-i lehe külastust. Samuti kirjeldatakse kasutatud tööriistu ning tehnikaid.
Marek Lepik. - B10 Interaktiivne mäng "Ereda leegiga" - WWW, PDF
Aparaaditehase "kultuuriplatvormi" raames loodud interaktiivne mäng, mille tarbeks on kombineeritud juba olemas olevaid IT süsteeme ja modifitseeritud tervikuks selliselt, et tekiks publiku ja etenduskunstniku vaheline visuaalne kommunikatsioon. Esitlusel on kohal ka etenduskunstnik, kes seda demonstreerib.
Ilo Tuule Rajand.
Magistri projektid:
- M01 ProtectWise - WWW, PDF
ProtectWise, the data-driven safety management solution that transforms workplace accident prevention. By leveraging advanced machine learning and a vast occupational accident dataset, it provides real-time risk assessments for individuals and teams. With ProtectWise, safety managers can proactively address hazards and promote a secure, productive work environment.
Mario Käära. - M02 Minesweeper With Friends - WWW, PDF
Minesweeper With Friends is a multiplayer 3D implementation of minesweeper, where players can solve the minefield together or compete with each other.
Kaarel Rüüsak. - M03 Harvest Mate - WWW, PNG
Harvest Mate is a smart planter that helps growing micro-greens at home for human consumption. The planter is embedded with sensors measuring the PH, ambient and soil humidity, temperature and light intensity. The user is notified when the plants need water and from data we wish to predict when to plant and harvest the greens.
Gautier Herve Jacques Reynes, Maksym Avramenko, Illia Tsiporenko, Oliver Vainikko. - M04 An Open-Source Robotic Study Companion For University Students - WWW, PDF
This project focused on addressing the gap in social robots designed for university students by developing an open-source robotic study companion (RSC) platform. Through extensive background research, valuable insights were gathered from existing educational solutions, which aided in carefully formulating the requirements and guiding the design and development of the RSC. A prototype was constructed using off-the-shelf components and leveraging OpenAI's conversational API, which has demonstrated its potential to simplify complex concepts for students.
Farnaz Baksh. - M05 SwinIA: Self-Supervised Blind-Spot Image Denoising without Convolutions - WWW, PDF
The first model to prove the concept of a pure-transformer blind-spot image denoising network with only one forward pass, simple MSE loss, and no demand for a known noise model.
Pavel Chizhov. - M06 Towards Decentralized Proof-of-Location - WWW, PDF
Making use of mesh network technologies and permissionless consensus mechanisms, we present a novel decentralized Proof-of-Location protocol and the implementation of a proof-of-concept, showing the achievement of space and time synchronization and the generation of complete, verifiable, and spatio-temporally sound location proofs.
Eduardo Ribas Brito. - M07 Predicting NMR spectra using low amounts of data - WWW, PDF
Nuclear magnetic resonance (NMR) spectra are the fingerprints of the molecule that can be used to identify the molecule's structure. Reading the spectrum requires specialized knowledge, and even then the decision often comes down to some potential candidates. Databases exist that collect NMR spectra so that the specialist can, through process of elimination, remove non-matching candidates. However, since the chemical space is vast and the machines required for measuring shifts are relatively expensive, then there is a need for accurate models that can predict spectra. The goal of this project was to develop model that could predict shifts using low amounts of data.
Herman Rull. - M08 eLex: AI-Powered Legal and Intellectual Property Platform - WWW, PDF
eLex is an AI-powered platform that can help startup businesses stay up-to-date with changing regulations by monitoring legal developments and providing actionable insights. This could include recommending changes to business processes or identifying potential compliance risks. Also, a platform can help businesses manage their intellectual property by automatically identifying and tracking patents, trademarks, and copyrights. This could also include monitoring for potential infringement and notifying the company of any potential threats.
Oleksandr Hranat, Yelyzaveta Denysenko. - M09 Reducing the Effect of Incomplete Annotations in Object Detection for Histopathology - WWW, PDF
Training neural networks for object detection usually requires decent amounts of data to produce great results. Apart from the image variety, the number of annotated objects is a crucial factor for success. In histopathology, the average annotation density is very high, resulting in resource-consuming data preparation for neural network training. We explore the effect of incomplete annotations in object detection. We show that modern object detectors, such as YOLO-v5, can effectively learn from histopathology datasets that lack up to 90% of annotations. Additionally, we suggest an easy model tuning setup to reduce the impact of incomplete annotations and enhance learning capability.
Denys Kaliuzhnyi, Mikhail Papkov. - M10 Shape Grammar Editor - WWW, PDF
This editor allows users to define shape grammar rules. Then these rules can be used to create 3D meshes. Additionally, the rules can be exported and used in Godot games to generate meshes during the run-time.
Mathias Plans. - M11 Veebirakendus programmeerimise õppimiseks II ja III kooliastmes - WWW, PDF
Programmeerimisõpik, mis on suunatud II ja III kooliastmele ning õpetab programmeerimist keeles Python.
Kristiina Keps. - M12 AirTech - WWW, PDF
Air Quality Monitoring System.
Nataliia Pabat, Serhii Yershykhin. - M13 XR Teleoperation Demo Development - WWW, PDF
Teleoperation is an important area of research in robotics and autonomous driving, but it can be challenging for students to understand and appreciate its potential without hands-on experience. To address this, RoverXR was developed to demonstrate key principles and capabilities of direct visual teleoperation in a clear and engaging way.
Matevž Borjan Zorec. - M14 "FiBar": a Tool for Automated Analysis of Complex Biomaterials from Microscopy Images - WWW, PDF
The goal of "FiBar" is to provide an automated microscopy image analysis tool for complex biomaterials. These complex biomaterials consist of fibers with encapsulated components, such as bacteria. The tool is divided into two pipelines - one for fiber diameter measuring, the other for bacterial cell analysis. The ultimate goal of this "FiBar" is to reduce the workload of the researcher by providing a viable alternative to the manual analysis.
Marilin Moor. - M15 Parameter-Efficient Fine-Tuning in Reading Comprehension - WWW, PDF
Analysing whether Parameter-Efficient Fine-Tuning can give better generalisation over fine-tuning in reading comprehension task.
Rustam Abdumalikov. - M16 Deblurring of microscopic 3D spheroid images using GANs - WWW, PDF
This project describes the approaches and experiments to deblur microscopic 3D images of spheroids using supervised and unsupervised neural networks.
Denys Krupovych.
