PhD student at Stony Brook University, NY. Previously a research assistant at the Aristotle University of Thessaloniki. Contact me at firstname.lastname@example.org.
We consider the problem of inferring high-dimensional data in a model that consists of a prior and an auxiliary constraint. The prior is an independently trained DDPM and the auxiliary constraint is expected to have a differentiable form but can come from diverse sources. The possibility of such inference turns diffusion models into plug-and-play modules, thereby allowing a range of potential applications, such as conditional generation, image segmentation and solving combinatorial optimization problems.
Internship project at Microsoft Research. Introduced a method for inferring labels across a collection of data samples where instead of hard labels we only have access to coarse and imprecise sources of information for each sample. The proposed algorithms were demonstrated on diverse problems ranging from classification with negative training examples, to co-segmentation of video frames and coarsely supervised text classification.
Part of the EU-funded research project Protein. Applying deep learning models to estimate the food volume in a given RGB image. Aiming to establish a user-friendly and easy-to-train system, able to accurately estimate the depicted food portion's nutritional value.
A novel approach to monocular depth estimation using Generative Adversarial Networks. The developed model was trained on the KITTI dataset where it exhibited promising results, showing the benefits of GANs versus common generative models.