Thesis offers

You are here

Check offers of other studies and specializations

Data Science and Computational Intelligence Knowledge Engineering and Machine Learning Hot Topics in AI and Professional Practice

Traditional Machine Learning systems require lots of training data to learn a particular task, and they are usually resilient to some noise in the data. But Large Language Models (LLM) such as GPT-3 or BLOOM are already trained in zillions of data, and much less data are needed to fine-tune them for a specific task. However, these data must be highly consistent and noise-free to ensure proper learning of the task. The project consist of devising and implentig a tool able to assists data curators in the creation, maintenance, and consistency checking of these datasets.

This project aims to analyze the prediction capability of Optical Coherence Tomography Angiography (OCTA) images for Diabetes Mellitus (DM) and Diabetic Retinopathy (DR,) in a large high-quality image dataset from previous research projects carried out in the field of Ophthalmology (Fundacio¿ La Marato¿ TV3, Fondo Investigaciones Sanitarias, FIS). OCTA is a newly developed, non-invasive, retinal imaging technique that permits adequate delineation of the perifoveal vascular network. It allows the detection of paramacular areas of capillary non perfusion and/or enlargement of the foveal avascular zone (FAZ), representing an excellent tool for assessment of DR.

Data Science and Computational Intelligence Vision, Perception and Robotics. Assistive Technologies Hot Topics in AI and Professional Practice

Explore the sate-of-the-art methods for (kinematic, multiperson) 3D human pose estimation from monocular video. Evaluate their performance when applied to strips of motion-picture films. Identify which approaches perform better and study their strenghs and limitations.

Data Science and Computational Intelligence Vision, Perception and Robotics. Assistive Technologies Hot Topics in AI and Professional Practice

Apply diffusion-based image generative models to convert videos into a cartoon style (e.g. from a sample image or a descriptive text). Depending on the obtained results a dictionary of styles may be created.

The student will have to implement different learning algorithms of Restricted Boltzmann Machine (RBM) neural networks using CUDA, and compare the performance against a standard CPU implementation.

Check offers of other studies and specializations