Semi-supervised Segmentation of Salt Bodies in Seismic Images Using an Ensemble of Convolutional Neural Networks

Abstract

Seismic image analysis plays a crucial role in a wide range of industrial applications and has been receiving significant attention. One of the essential challenges of seismic imaging is detecting subsurface salt structure which is indispensable for the identification of hydrocarbon reservoirs and drill path planning. Unfortunately, the exact identification of large salt deposits is notoriously difficult and professional seismic imaging often requires expert human interpretation of salt bodies. Convolutional neural networks (CNNs) have been successfully applied in many fields, and several attempts have been made in the field of seismic imaging. But the high cost of manual annotations by geophysics experts and scarce publicly available labeled datasets hinder the performance of the existing CNN-based methods. In this work, we propose a semi-supervised method for segmentation (delineation) of salt bodies in seismic images which utilizes unlabeled data for multi-round self-training. To reduce error amplification during self-training we propose a scheme which uses an ensemble of CNNs. We show that our approach outperforms state-of-theart on the TGS Salt Identification Challenge dataset and is ranked the first among the 3234 competing methods. The source code is available at GitHub.

Identification of Salt Deposits on Seismic Images Using Deep Learning Method for Semantic Segmentation

“…Along with private scores that were the basis for the final ranking, corresponding public scores, that were visible during the competition, are also shown. For comparison purposes, the winning solution [41] score is also shown. The total number of submitted solutions for the winning team was 316, while the author had 42 total submissions.…”

“…Motivated by good results in many areas, CNNs are becoming researchers' default choice for the segmentation of seismic images and identification of salt deposits. A huge number of papers [34][35][36][37][38][39][40][41] in 2018 and 2019 supports the claim. Dramsch and Lüthje [34] evaluated several classification deep CNNs with transfer learning to identify nine different seismic textures from 65 × 65 pixel patches.…”

“…Wu et al [40] extended this approach to 3D and directly processed 128 × 128 × 128 data cubes to predict salt bodies. Finally, a paper by Babakhin et al [41] should be emphasized since it describes the first-place solution on the TGS's challenge [2]. Their key 'ingredient to success' was a semi-supervised method which utilizes unlabeled (test) data for multi-round self-training.…”

“…The prediction is performed on 2D images instead of approaches [37,40] that process 3D data cubes. It is based on a novel network architecture, that is presented in the next section, which distinguishes it from [35,38,39,41].…”

Fast and Efficient Model for Real-Time Tiger Detection In The Wild

“…The model has 600k parameters, requires 0.071 GFLOPs per image and can run on the edge devices (smart cameras) in near real time. In addition, we introduce a two-stage semi-supervised learning via pseudo-labelling learning approach to distill the knowledge from the larger networks [12] [2]. For ATRW-ICCV 2019 tiger detection sub-challenge, based on public leaderboard score, our approach shows superior performance in comparison to other methods.…”