English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Geometric transformations have been widely used to augment the size oftraining images. Existing methods often assume a unimodal distribution of theunderlying transformations between images, which limits their power when datawith multimodal distributions occur. In this paper, we propose a novel model,Multimodal Geometric Augmentation (MGAug), that for the first time generatesaugmenting transformations in a multimodal latent space of geometricdeformations. To achieve this, we first develop a deep network that embeds thelearning of latent geometric spaces of diffeomorphic transformations (a.k.a.diffeomorphisms) in a variational autoencoder (VAE). A mixture of multivariateGaussians is formulated in the tangent space of diffeomorphisms and serves as aprior to approximate the hidden distribution of image transformations. We thenaugment the original training dataset by deforming images using randomlysampled transformations from the learned multimodal latent space of VAE. Tovalidate the efficiency of our model, we jointly learn the augmentationstrategy with two distinct domain-specific tasks: multi-class classification on2D synthetic datasets and segmentation on real 3D brain magnetic resonanceimages (MRIs). We also compare MGAug with state-of-the-art transformation-basedimage augmentation algorithms. Experimental results show that our proposedapproach outperforms all baselines by significantly improved predictionaccuracy. Our code is publicly available athttps://github.com/tonmoy-hossain/MGAug.

MGAug: Multimodal Geometric Augmentation in Latent Spaces of Image Deformations