A list of some of the Projects I have worked/ working on independently or as a part of my Master’s program
Improving object representations in low light conditions through unpaired GAN based image enhancement:
In this project, an annotation-free method to enhance images at different ranges of luminance is explored using Generative Adversarial Networks.Thecruxof this project lies in utilizing an object-context based unsupervised method to train EnlightenGAN in contrast to the method from the paper.
Implemented as a part of Computer Vision KEN4255 Final exam.
Analysis and Review of Stochastic Weight Averaging Gaussian
Stochastic Weight Averaging – Gaussian, an approach for uncertainty representation and calibration is reproduced, reviewed and evaluated for computationally efficient Bayesian inference in deep learning. This approach allows representing uncertainty over a model’s decisions which can be critical in domains such as medical or legal decision making.