Self-supervised noise suppression¶

Instructors¶

• Claire Birnie - Seismic Wave Analysis Group, KAUST
• Sixie Liu - Seismic Wave Analysis Group, KAUST

Tutorial Objectives¶

Self-supervised learning offers a solution to the common limitation of the lack of noisy-clean pairs of data for training deep learning seismic denoising procedures.

In this tutorial, we will explain the theory behind blind-spot networks and how these can be used in a self-supervised manner, removing any requirement of clean-noisy training data pairs. We will deep dive into how the original methodologies for random noise can be adapted to handle realistic noise in seismic data, both pseudo-random noise, and structured noise. Furthermore, each sub-topic presented will be followed by a live, code-along session such that all participants will be able to recreate the work shown and can afterward apply it to their own use cases.

What you’ll need:¶

• Slack channel: #t22-wed-noise-suppression
• Dataset:
• Course materials: