Lung VAE

Lung VAE, and more broadly, Variational Autoencoders (VAEs), are probabilistic generative models used for learning data representations and generating new data samples. Current research focuses on improving VAE performance through architectural modifications (e.g., incorporating transformers, incorporating frequency information, and using different prior distributions like Student's t-distributions), and addressing limitations such as blurry reconstructions and posterior collapse. These advancements have implications across diverse fields, including medical image analysis (e.g., lung segmentation, craniofacial syndrome diagnosis), drug discovery (molecule generation), and time series anomaly detection, by enabling more accurate and interpretable models.