Deep-learning seismic facies on state-of-the-art CNN architectures

Jesper Sören Dramsch; Mikael Lüthje

doi:10.1190/segam2018-2996783.1

Deep-learning seismic facies on state-of-the-art CNN architectures

Jesper Sören Dramsch, Mikael Lüthje

Department of Applied Mathematics and Computer Science

Research output: Contribution to conference › Paper › Research › peer-review

Abstract

We explore propagation of seismic interpretation by deep learning in stacked 2D sections. We show the application of state-of-the-art image classification algorithms on seismic data. These algorithms were trained on big labeled photograph databases. We use transfer learning to benefit from pre-trained networks and evaluate their performance on seismic data.

Original language	English
Publication date	2018
DOIs	https://doi.org/10.1190/segam2018-2996783.1
Publication status	Published - 2018
Event	SEG International Exposition and 88th Annual Meeting - Anaheim, United States Duration: 14 Oct 2018 → 19 Oct 2018 Conference number: 88

Conference

Conference	SEG International Exposition and 88th Annual Meeting
Number	88
Country/Territory	United States
City	Anaheim
Period	14/10/2018 → 19/10/2018

Access to Document

10.1190/segam2018-2996783.1

OpenUrl availability

Full text

Cite this

@conference{0c9e7a201b2a46a6a2bc20f09ed2295f,

title = "Deep-learning seismic facies on state-of-the-art CNN architectures",

abstract = "We explore propagation of seismic interpretation by deep learning in stacked 2D sections. We show the application of state-of-the-art image classification algorithms on seismic data. These algorithms were trained on big labeled photograph databases. We use transfer learning to benefit from pre-trained networks and evaluate their performance on seismic data. ",

author = "\{S{\"o}ren Dramsch\}, Jesper and Mikael L{\"u}thje",

year = "2018",

doi = "10.1190/segam2018-2996783.1",

language = "English",

pages = "2036--2040",

note = "SEG International Exposition and 88th Annual Meeting, SEG18 ; Conference date: 14-10-2018 Through 19-10-2018",

}

TY - CONF

T1 - Deep-learning seismic facies on state-of-the-art CNN architectures

AU - Sören Dramsch, Jesper

AU - Lüthje, Mikael

N1 - Conference code: 88

PY - 2018

Y1 - 2018

N2 - We explore propagation of seismic interpretation by deep learning in stacked 2D sections. We show the application of state-of-the-art image classification algorithms on seismic data. These algorithms were trained on big labeled photograph databases. We use transfer learning to benefit from pre-trained networks and evaluate their performance on seismic data.

AB - We explore propagation of seismic interpretation by deep learning in stacked 2D sections. We show the application of state-of-the-art image classification algorithms on seismic data. These algorithms were trained on big labeled photograph databases. We use transfer learning to benefit from pre-trained networks and evaluate their performance on seismic data.

U2 - 10.1190/segam2018-2996783.1

DO - 10.1190/segam2018-2996783.1

M3 - Paper

SP - 2036

EP - 2040

T2 - SEG International Exposition and 88th Annual Meeting

Y2 - 14 October 2018 through 19 October 2018

ER -

Deep-learning seismic facies on state-of-the-art CNN architectures

Abstract

Conference

Access to Document

OpenUrl availability

Fingerprint

Cite this