VAESim: A probabilistic approach for self-supervised prototype discovery

In medical image datasets, discrete labels are often used to describe a continuous spectrum of conditions, making unsupervised image stratification a challenging task. In this work, we propose VAESim, an architecture for image stratification based on a conditional variational autoencoder. VAESim lea...

Full description

Saved in:
Bibliographic Details
Published in:Image and vision computing Vol. 137; p. 104746
Main Authors: Ferrante, Matteo, Boccato, Tommaso, Spasov, Simeon, Duggento, Andrea, Toschi, Nicola
Format: Journal Article
Language:English
Published: Elsevier B.V 01.09.2023
Subjects:
Deep clustering
Medical imaging
Prototypes discovery
Variational autoencoders
Variational autoencoders
Medical imaging
Prototypes discovery
Deep clustering
ISSN:0262-8856
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In medical image datasets, discrete labels are often used to describe a continuous spectrum of conditions, making unsupervised image stratification a challenging task. In this work, we propose VAESim, an architecture for image stratification based on a conditional variational autoencoder. VAESim learns a set of prototypical vectors during training, each associated with a cluster in a continuous latent space. We perform a soft assignment of each data sample to the clusters and reconstruct the sample based on a similarity measure between the sample embedding and the prototypical vectors. To update the prototypical embeddings, we use an exponential moving average of the most similar representations between actual prototypes and samples in the batch size. We test our approach on the MNIST handwritten digit dataset and the PneumoniaMNIST medical benchmark dataset, where we show that our method outperforms baselines in terms of kNN accuracy (up to +15% improvement in performance) and performs at par with classification models trained in a fully supervised way. Our model also outperforms current end-to-end models for unsupervised stratification. [Display omitted] •Discover prototypes and clusters in image data with VAE's new prototypes matrix.•Improve VAE's latent space for clustering, sampling, and downstream tasks.•Novel and simple approach, outperforms other deep clustering algorithms•Self-supervised training enables data generation, reconstruction, and learning•The model is benchmarked on famous images dataset, including a medical dataset.
ISSN:0262-8856
DOI:10.1016/j.imavis.2023.104746