The first DeepCube scientific paper “Self-Supervised Contrastive Learning for Volcanic Unrest Detection” is published in IEEE Geoscience and Remote Sensing Letters.

Our team conducting research on volcanic unrest detection and alerting, introduces a self-supervised learning framework for volcanic activity detection globally.

Proposed pipeline. First, we use unlabeled InSAR data to learn feature representations with the SimCLR self-supervised framework and then attach a linear classifier for the supervised training with a few labels.

We show that our self-supervised pipeline achieves higher accuracy with respect to the state-of-the-art methods and shows excellent generalization even for out-of-distribution test data. We showcase the effectiveness of our approach for detecting the unrest episodes preceding the recent Icelandic Fagradalsfjall volcanic eruption.

Visualization of ResNet50 activations on Fagradalsfjall volcano. Pink represents the area that affected the network’s decision the most. Both unrest episodes are shown in chronological order. PrSimCLR and expert are the predictions made by our method and the InSAR expert, respectively, (1=positive and 0=negative).

To access and download the publication visit

Cite the publication: N. I. Bountos, I. Papoutsis, D. Michail and N. Anantrasirichai, “Self-Supervised Contrastive Learning for Volcanic Unrest Detection,” in IEEE Geoscience and Remote Sensing Letters, doi: 10.1109/LGRS.2021.3104506.