Semantic Segmentation of Seismic Reflection

The project code is available on Github.

Title imageExample of a whole seismic mosaic post-processing with salt domes. Green/blue - salt/empty regions from the train dataset; red - predicted mask; yellow - inpainted by the post-processing.

The project is based on the Kaggle TGS Salt Identification Challenge competition, which goal was to build an algorithm that automatically and accurately identifies if a subsurface target is a salt deposit or not on seismic images. The task is crucial for oil and gas company drillers. The object of the competition is seismic data collected using reflection seismology. In a nutshell, the problem can be formulated as a semantic segmentation computer vision task.

Detailed description of our solution is available here.

© 2019-2024 OniroAI (former ClubAI)