Computer Vision and Machine Learning

This study explores the use of deep learning, specifically CycleGAN, to synthesize angiographic fundus images from conventional color fundus images, aiming to provide a safer alternative to traditional angiographic imaging which carries potential risks.

This study explores the use of deep learning, specifically CycleGAN, to synthesize angiographic fundus images from conventional color fundus images, aiming to provide a safer alternative to traditional angiographic imaging which carries potential risks.

Our study introduces SeLFVi, a self-supervised algorithm for reconstructing high-fidelity light field videos from stereo video, utilizing geometric and temporal data from stereo pairs and applying a low-rank constraint for light field display, significantly enhancing post-capture video processing on mobile devices.

The paper introduces medXGAN, a generative adversarial framework designed to visually explain medical classifiers’ decisions by disentangling anatomical structures and pathologies, offering improved localization and explanatory abilities over existing methods like Grad-CAM and Integrated Gradients.