object extraction from satellite imagery using deep learning

27
Object Extraction from Satellite Imagery using Deep Learning Aly Osama

Upload: aly-osama

Post on 21-Jan-2018

1.963 views

Category:

Education


1 download

TRANSCRIPT

Page 1: Object extraction from satellite imagery using deep learning

Object Extraction from Satellite Imagery using Deep Learning

Aly Osama

Page 2: Object extraction from satellite imagery using deep learning

Presenter

• Aly Osama

• Research Software Development Engineer at Microsoft

• Contact:

• Email: [email protected]

• https://www.linkedin.com/in/alyosama/

• https://github.com/alyosama

Page 3: Object extraction from satellite imagery using deep learning

Agenda

1. Needed data, its size,

2. How training will go,

3. How evaluation should be carried,

4. What learning tools would you see and why?

5. Literature survey on the subject

6. Code sample/preliminary results

Page 4: Object extraction from satellite imagery using deep learning

Overview

Key Words:-

1. Satellite Imagery 2. Object Extraction3. Deep Learning4. Python

Page 5: Object extraction from satellite imagery using deep learning

1. Satellite Imagery

• Objects are often very small (~20 pixels in size ) as example 0.5m/pixel

• Input images are enormous (often hundreds of megapixels).

• Image has more than 3 channels (RGB) called bands.

• Image Format:• Images (GeoTiff , … etc )• Labels ( GeoJSON , WTK )

• On the positive side, • The physical and pixel scale of objects

are usually known in advance• There’s a low variation in observation

angle.

Page 6: Object extraction from satellite imagery using deep learning

Datasets • SpaceNet

• Source : https://aws.amazon.com/public-datasets/spacenet/

• Tutorial https://medium.com/the-downlinq/building-extraction-with-yolt2-and-spacenet-data-a926f9ffac4f

• 5 Solutions: https://github.com/SpaceNetChallenge/BuildingDetectors

• DSTL Satellite Imagery (Kaggle Competition )• https://www.kaggle.com/c/dstl-satellite-imagery-

feature-detection• Tutorial : https://www.kaggle.com/torrinos/dstl-

satellite-imagery-feature-detection/exploration-and-plotting/run/553107

• DeepSat

• UC Merced land

• Satellite Websites• https://modis.gsfc.nasa.gov/• https://sentinel.esa.int/web/sentinel/home• https://landsat.usgs.gov/

Page 7: Object extraction from satellite imagery using deep learning

2. Training

• Experiments

1. VGGNet. - (Baseline)• Tune the pretrained model – Transfer

Learning – based on available data

• Data Augmentation like • Random Crops / Scales

• Color Jitter

2. Faster RCNN or YOLO• For Detection and Localization

Page 8: Object extraction from satellite imagery using deep learning

General Tips

Page 9: Object extraction from satellite imagery using deep learning

3.Evaluation

• Accuracy :D

• Precession and Recall

• Jaccard index ( Intersection over Union)

Page 10: Object extraction from satellite imagery using deep learning

4. tools

• Hardware • Deep Learning AMI Amazon Linux Version

• https://aws.amazon.com/marketplace/pp/B01M0AXXQB• Powerful Machine with GPUs

• NVIDIA GTX Titan X

• Software• Caffe + Python

• Pretrained models • Lasagne / Keras Framework

• High level Python• Backend independent “Tensorflow or Theano” • Multi-GPU

• Caffe2 ( 18-4-2017 )

• Utility• QGIS is an open-source tool for managing and editing GeoTIFFs and

geoJSON files

Page 11: Object extraction from satellite imagery using deep learning

Tools Comparison

Page 12: Object extraction from satellite imagery using deep learning

5. Literature survey on the subject

Page 13: Object extraction from satellite imagery using deep learning

1- A Survey on Object Detection in Optical Remote Sensing Images 2016Gong Cheng, Junwei Han*

Page 14: Object extraction from satellite imagery using deep learning

Cheng et al. 2016

Page 15: Object extraction from satellite imagery using deep learning

Cheng et al. 2016

• Deep Learning Papers • (Han et al., 2015;Tang et al.,

2015; Wang et al., 2015; Zhou et al., 2015b)

• Datasets• NWPU VHR-10 dataset (Cheng et

al., 2014a)1. • SZTAKI-INRIA building detection

dataset (Benedek et al., 2012)2. • TAS aerial car detection dataset

(Heitz and Koller, 2008)3.• Overhead imagery research

dataset (OIRDS) (Tanner et al., 2009)

• IITM road extraction dataset (Das et al., 2011)5.

Page 16: Object extraction from satellite imagery using deep learning

FAST AIRCRAFT DETECTION IN SATELLITE IMAGES BASED ON CONVOLUTIONAL NEURAL NETWORKS (Wu et. Al.) 2015

Page 17: Object extraction from satellite imagery using deep learning

Region Proposals

TODO : Search on R-CNN

Page 18: Object extraction from satellite imagery using deep learning

FULLY CONVOLUTIONAL NETWORKS FOR BUILDING AND ROAD EXTRACTION: PRELIMINARY RESULTS 2016 Zilong Zhong

• Dataset• Massachusetts’ road dataset and building dataset • each image consists of 3×1500×1500 pixels• contains 1,711 aerial images,

• the FCN’s computation consumption could be much higher than that of the ordinary object recognition models.

Page 19: Object extraction from satellite imagery using deep learning

Building detection in very high resolution multispectral data with deep learning features 2016• AlexNet -> Features + SVM ( Last layer )

Page 20: Object extraction from satellite imagery using deep learning

Road network extraction a neural-dynamic framework based on deep learning and a finite state machine wang2015

CNN + FSM ( for sequence )

Page 21: Object extraction from satellite imagery using deep learning

Do Deep Features Generalize from Everyday Objectsto Remote Sensing and Aerial Scenes Domains? 2016Ot´avio A. B. Penatti

• ConvNet using Caffe and OverFeat

Page 22: Object extraction from satellite imagery using deep learning

Using convolutional networks and satellite imagery to identify pa.erns inurban environments at a large scaleADRIAN ALBERT*, Massachuse.s Institute of Technology2017

• Dataset :

• UC Merced land use dataset [25] (of 2100 images spanning 21 classes)

• DeepSat land use benchmark dataset ( 4 channels

• (VGGNet and ResNet)

Page 23: Object extraction from satellite imagery using deep learning

6. Code sample/prelim

inary results

• (Test) DSTL

• https://www.kaggle.com/alyosama/dstl-satellite-imagery-feature-detection/convnet-baseline/

Page 24: Object extraction from satellite imagery using deep learning

YOLT2

• The actual F1 score of 0.21

• Jaccard index between 0.4 and 0.5

https://medium.com/the-downlinq/building-extraction-with-yolt2-and-spacenet-data-a926f9ffac4fT

Page 25: Object extraction from satellite imagery using deep learning

CosmiQNet

• Blackbox Fully Convolution Neural Network: CosmiQNet. The inputs are at two resolutions and the output distance transform matches the lower of the input resolutions. The resolution of the 8-band GeoTIFF is roughly one quarter (in each dimension) the resolution of the the 3-band GeoTIFF; the difference in resolution is depicted by the scale of the GeoTIFFs.

https://medium.com/the-downlinq/object-detection-on-spacenet-5e691961d257

Page 26: Object extraction from satellite imagery using deep learning

Resources

1. DSTL Satellite Imagery Competiton

2. https://medium.com/@avanetten

3. https://www.kernix.com/blog/image-classification-with-a-pre-trained-deep-neural-network_p11

Page 27: Object extraction from satellite imagery using deep learning

[email protected]

Thank you