book

Deep Learning for Remote Sensing Images with Open Source Software

Name: Deep Learning for Remote Sensing Images with Open Source Software
Author: Rémi Cresson
ISBN: 9781000093612

by Rémi Cresson

July 2020

Intermediate to advanced

164 pages

3h 16m

English

CRC Press

Read now

Unlock full access

Cover
Half Title
Series Page
Title Page
Copyright Page
Contents
Preface
Author
Part I: Backgrounds
1. Deep learning background
1.1 What is deep learning?1.2 Convolution 1.3 Pooling1.4 Activation functions 1.5 Challenges ahead for deep learning with remote sensing images

2. Software
2.1 Orfeo ToolBox2.1.1 Applications2.1.2 Streaming mechanism2.1.3 Remote modules2.1.4 The Python API 2.2 TensorFlow2.2.1 APIs2.2.2 Computations2.2.3 Graphs2.3 Orfeo ToolBox + TensorFlow = OTBTF2.3.1 Installation 2.3.2 Featured applications2.3.3 Principle2.3.4 Multiple input sources and outputs2.4 QGIS
Part II: Patch-based classification
3. Introduction
4. Data used: the Tokyo dataset
4.1 Description4.2 Remote sensing imagery 4.3 Terrain truth
5. A simple convolutional neural network
5.1 Normalization5.2 Sampling5.2.1 Selection5.2.2 Extraction5.3 Training5.3.1 Principle5.3.2 Model architecture5.3.2.1 Input5.3.2.2 Layers5.3.2.3 Estimated class5.3.2.4 Loss function5.3.2.5 Optimizer5.4 Generate the model5.5 Train the model from scratch5.6 Comparison with Random Forest5.7 Inference
6. Fully Convolutional Neural Network
6.1 Using the existing model as an FCN 6.2 Pixel-wise fully convolutional model6.3 Training6.4 Inference
7. Classifiers on deep features
7.1 Principle 7.2 Overview of composite applications in OTB7.3 Training 7.4 Inference
8. Dealing with multiple sources
8.1 More sources?8.2 Model with multiple inputs 8.3 Normalization 8.4 Sampling8.5 Training8.5.1 Inference8.5.1.1 Patch-based mode8.5.1.2 Fully convolutional mode
9. Discussion
Part III: Semantic segmentation
10. Semantic segmentation of optical imagery
10.1 Introduction10.2 Overview
11. Data used: the Amsterdam dataset
11.1 Description 11.2 Spot-7 image11.3 OpenStreetMap data11.3.1 OSM downloader plugin11.3.2 Download OSM data11.3.3 Prepare the vector layer
12. Mapping buildings
12.1 Input data pre-processing12.1.1 Satellite image pansharpening 12.1.2 Image normalization 12.1.3 Sample selection12.1.3.1 Patch position seeding 12.1.3.2 Patch position selection12.1.3.3 Patches split12.1.4 Rasterization12.1.5 Patch extraction12.2 Building the model12.2.1 Architecture12.2.2 Implementation12.2.2.1 Exact output12.2.2.2 Expression field 12.2.3 Generate the SavedModel12.3 Training the model12.4 Inference
13. Discussion
Part IV: Image restoration
14. Gapfilling of optical images: principle
14.1 Introduction 14.2 Method14.3 Architecture14.3.1 Encoder 14.3.2 Decoder14.3.3 Loss
15. The Marmande dataset
15.1 Description15.2 Sentinel-2 images15.3 Sentinel-1 image
16. Pre-processing
16.1 Sentinel images16.1.1 Optical images16.1.2 SAR image16.1.2.1 Calibration16.1.2.2 Filtering values16.1.2.3 Linear stretch16.1.2.4 Spatial resampling 16.2 Patches16.2.1 Patch position seeding16.2.1.1 Sentinel-2 image masks16.2.1.2 Merge masks16.2.1.3 Grid generation16.2.1.4 Grid filtering16.2.1.5 Patch centroids16.2.1.6 Training and validation datasets 16.2.2 Extraction of patches16.3 More: automate steps with the OTB Python API16.3.1 Build the pipeline16.3.2 Run the pipeline
17. Model training
17.1 Training from Python17.2 Get the code17.3 Use the code17.3.1 Description17.3.2 Parameters17.4 Export the model
18. Inference
18.1 Inputs and outputs18.2 Generating the image18.3 Postprocessing
19. Discussion
Bibliography
Index

Content preview from Deep Learning for Remote Sensing Images with Open Source Software

A simple convolutional neural network

We will start with something easy: a small convolutional neural network. First, we will look at the data that will be employed to train and use our network. Open QGIS and import the geospatial data. The raster files can be imported from the Layer menu as shown in figure 5.1.

fig5_1 — **FIGURE 5.1**Open raster file.

Prepare a directory for the results somewhere on your file system, where you have the write permission accorded. We encourage the reader to use this directory to store all resulting files, and to run all command lines from inside this directory.

Use your favorite interface

The document shows how ...

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.

Read now

Unlock full access

More than 5,000 organizations count on O’Reilly

O’Reilly covers everything we've got, with content to help us build a world-class technology community, upgrade the capabilities and competencies of our teams, and improve overall team performance as well as their engagement.

Julian F.

Head of Cybersecurity

I wanted to learn C and C++, but it didn't click for me until I picked up an O'Reilly book. When I went on the O’Reilly platform, I was astonished to find all the books there, plus live events and sandboxes so you could play around with the technology.

Addison B.

Field Engineer

I’ve been on the O’Reilly platform for more than eight years. I use a couple of learning platforms, but I'm on O'Reilly more than anybody else. When you're there, you start learning. I'm never disappointed.

Amir M.

Data Platform Tech Lead

I'm always learning. So when I got on to O'Reilly, I was like a kid in a candy store. There are playlists. There are answers. There's on-demand training. It's worth its weight in gold, in terms of what it allows me to do.

Mark W.

Embedded Software Engineer

Subpixel Mapping for Remote Sensing Images

Publisher Resources

ISBN: 9781000093612

Cloud Computing

Data Engineering

Data Science

AI & ML

Programming Languages

Software Architecture

IT/Ops

Security

Design

Business

Soft Skills

Deep Learning for Remote Sensing Images with Open Source Software

by Rémi Cresson

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.