SlideShare a Scribd company logo
1 of 25
TRAINING DEEP LEARNING MODELS TO COUNT
USING SYNTHETIC IMAGES
6 Sept. 2019DR. ANDREAS KAMILARIS
@DL-UAV19, PROC. OF CAIP 2019
Problem
• Difficult to create ground truth data
• UCSD Pedestrian Database
• Video of pedestrians on UCSD walkways, taken
from a stationary camera
Problem
• Difficult to create ground truth data
• UCF CC 50 dataset
• Counts of persons range between 94 and 4543, with an average
of 1280 individuals per image
Motivation
• Generate data by simulation, which are easy then to label.
• Create synthetic ground truth data
Rahnemoonfar, M. and Sheppard, C., 2017. Deep count: fruit counting based on deep
simulated learning. Sensors, 17(4), p.905.
First try
• Generated simulated data in Python
• Goal: Detection of fires in forests
• 160 images, 80 of forest, 80 of fire
• 80% training, 20% testing
First try
• Model: Inception-v3 vs. Custom
First try
• False negatives where there was smoke AND fire!
Data Custom Inception-v3
Only
generative
0.65 0.62
Only real 0.8 0.69
Combined,
augmented
dataset
0.9 0.71
Second try
• Better representation of fire
Second try
• Improvement of results!
• Very few false negatives
Data Custom Inception-v3
Only
generative
0.95 0.72
Only real 0.8 0.69
Combined,
augmented
dataset
1.0 0.79
Generative data case
Training based on 2,000
synthetic images (labelled as
fire or forest)
Testing based on 100 real-world aerial
photos (classified as 50 images of
forest and 50 images of fire)
Hypothesis 1
Generating synthetic data can help to train deep learning models,
without the need to create expensive (in terms of time and effort)
ground truth data!
Hypothesis 2
Generating synthetic data can help to train deep learning models
not only to classify, but also to count!
… not only simple problems, but also more advanced ones…
Application: Counting houses from aerial photos
 60 photos taken from satellite images in urban areas of Tanzania
 Manually counted the number of houses, to create out
testing/validation dataset.
 Each photo has [0,38] houses.
Application: Counting houses from aerial photos
 Created synthetic training data, with automated house counting
 First naïve try: Involved only squares!
MSE = 41
Application: Counting houses from aerial photos
 Second try: Added trees and small shadows
MSE = 29
Application: Counting houses from aerial photos
 Third try: Added grass, fences and different orientation of
houses. Also added images without any houses
MSE = 20
Application: Counting houses from aerial photos
 Many decisions to take along the way…
• Dropout rate (35-50% works well)
• Stride (2 is small, 10 is too big)
• Convolutions (7x7 initially seems a good option)
• Pre-training (ImageNet is not helping a lot)
• Max-pooling better than average pooling
• Dense layers at the end of the network, ReLU function
Application: Counting houses from aerial photos
Adapted,
custom
topology
Inception-ResNet
7x7 input filter with
large stride
Dense fully-
connected layer
with ReLU
Application: Counting houses from aerial photos
 Training vs. Testing MSE
The model can predict the number of houses with an error of 4,47 houses.
For example, for a photo with 20 houses, the model would predict in the
range of [16, 24].
Training based on 10,000
synthetic images (labelled
with exact number of houses)
Testing based on 60 real-
world aerial photos (labelled
with exact number of houses)
Application: Counting houses from aerial photos
 Best Vs. Worst predictions
Application: Counting houses from aerial photos
 Next steps:
• Crop houses from training dataset and reuse based on
random combinations in semi-synthetic images
• More realistic generation of data
• (GAN for counting? )
• Accountability
• Other domains:
o Agriculture (counting animals in farms)
o Energy (renewable energy in roofs)
o Environment and Climate (counting trees, plants,
endangered species of animals etc.)
o Microbiology (blood test analysis etc.)
 State of the art (published in 2019)
Kar, A., Prakash, A., Liu, M.Y., Cameracci, E., Yuan, J., Rusiniak, M., Acuna, D., Torralba, A. and Fidler, S.,
2019. Meta-Sim: Learning to Generate Synthetic Datasets. arXiv preprint arXiv:1904.11621.
Future Work and Research Direction
Future Work and Research Direction
 State of the art (published in 2019)
 Combining Counting CNN model with the ResNeXt architecture
Tian, M. et al., Automated pig counting using deep learning,
Computers and Electronics in Agriculture, vol. 163, pp. 1-10, 2019. MAE = 4.47
MAE = 2.77
Conclusion
• Synthetic data can be used for training DL models
• Can be applied in UAV-related applications (classification vs.
counting problems)
• More advanced techniques are required for improving
performance (e.g. probabilistic scene graph generation, density
maps)
THANKS FOR YOUR ATTENTION!
DR. ANDREAS KAMILARIS
EMAIL: A.KAMILARIS@UTWENTE.NL

More Related Content

Similar to Training deep learning models to count using synthetic images

APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENTAPPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENTsipij
 
Artem Baklanov - Votes Aggregation Techniques in Geo-Wiki Crowdsourcing Game:...
Artem Baklanov - Votes Aggregation Techniques in Geo-Wiki Crowdsourcing Game:...Artem Baklanov - Votes Aggregation Techniques in Geo-Wiki Crowdsourcing Game:...
Artem Baklanov - Votes Aggregation Techniques in Geo-Wiki Crowdsourcing Game:...AIST
 
H2O with Erin LeDell at Portland R User Group
H2O with Erin LeDell at Portland R User GroupH2O with Erin LeDell at Portland R User Group
H2O with Erin LeDell at Portland R User GroupSri Ambati
 
Detecting solar farms with deep learning
Detecting solar farms with deep learningDetecting solar farms with deep learning
Detecting solar farms with deep learningJason Brown
 
Learning new climate science by thinking creatively with machine learning
Learning new climate science by thinking creatively with machine learningLearning new climate science by thinking creatively with machine learning
Learning new climate science by thinking creatively with machine learningZachary Labe
 
Neural Networks and Deep Learning for Physicists
Neural Networks and Deep Learning for PhysicistsNeural Networks and Deep Learning for Physicists
Neural Networks and Deep Learning for PhysicistsHéloïse Nonne
 
IBM Cloud Paris Meetup 20180517 - Deep Learning Challenges
IBM Cloud Paris Meetup 20180517 - Deep Learning ChallengesIBM Cloud Paris Meetup 20180517 - Deep Learning Challenges
IBM Cloud Paris Meetup 20180517 - Deep Learning ChallengesIBM France Lab
 
The effects of visual realism on search tasks in mixed reality simulations-IE...
The effects of visual realism on search tasks in mixed reality simulations-IE...The effects of visual realism on search tasks in mixed reality simulations-IE...
The effects of visual realism on search tasks in mixed reality simulations-IE...Yadhu Kiran
 
Himansu sahoo resume-ds
Himansu sahoo resume-dsHimansu sahoo resume-ds
Himansu sahoo resume-dsHimansu Sahoo
 
Practical computer vision-- A problem-driven approach towards learning CV/ML/DL
Practical computer vision-- A problem-driven approach towards learning CV/ML/DLPractical computer vision-- A problem-driven approach towards learning CV/ML/DL
Practical computer vision-- A problem-driven approach towards learning CV/ML/DLAlbert Y. C. Chen
 
TraitCapture: NextGen phenomics tools for lab and field [ComBio2015]
TraitCapture: NextGen phenomics tools for lab and field [ComBio2015]TraitCapture: NextGen phenomics tools for lab and field [ComBio2015]
TraitCapture: NextGen phenomics tools for lab and field [ComBio2015]TimeScience
 
Microstructural Analysis and Machine Learning
Microstructural Analysis and Machine LearningMicrostructural Analysis and Machine Learning
Microstructural Analysis and Machine LearningPFHub PFHub
 
TERN eMAST : Observations and terrestrial ecosystem models : Terrestrial Ecos...
TERN eMAST : Observations and terrestrial ecosystem models : Terrestrial Ecos...TERN eMAST : Observations and terrestrial ecosystem models : Terrestrial Ecos...
TERN eMAST : Observations and terrestrial ecosystem models : Terrestrial Ecos...Brad Evans
 
Machine Learning ebook.pdf
Machine Learning ebook.pdfMachine Learning ebook.pdf
Machine Learning ebook.pdfHODIT12
 
1_5_AI_edx_ml_51intro_240204_104838machine learning lecture 1
1_5_AI_edx_ml_51intro_240204_104838machine learning lecture 11_5_AI_edx_ml_51intro_240204_104838machine learning lecture 1
1_5_AI_edx_ml_51intro_240204_104838machine learning lecture 1MostafaHazemMostafaa
 
Machine learning in scientific workflows
Machine learning in scientific workflowsMachine learning in scientific workflows
Machine learning in scientific workflowsBalázs Kégl
 
Introduction to the Artificial Intelligence and Computer Vision revolution
Introduction to the Artificial Intelligence and Computer Vision revolutionIntroduction to the Artificial Intelligence and Computer Vision revolution
Introduction to the Artificial Intelligence and Computer Vision revolutionDarian Frajberg
 

Similar to Training deep learning models to count using synthetic images (20)

APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENTAPPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
 
Artem Baklanov - Votes Aggregation Techniques in Geo-Wiki Crowdsourcing Game:...
Artem Baklanov - Votes Aggregation Techniques in Geo-Wiki Crowdsourcing Game:...Artem Baklanov - Votes Aggregation Techniques in Geo-Wiki Crowdsourcing Game:...
Artem Baklanov - Votes Aggregation Techniques in Geo-Wiki Crowdsourcing Game:...
 
H2O with Erin LeDell at Portland R User Group
H2O with Erin LeDell at Portland R User GroupH2O with Erin LeDell at Portland R User Group
H2O with Erin LeDell at Portland R User Group
 
Detecting solar farms with deep learning
Detecting solar farms with deep learningDetecting solar farms with deep learning
Detecting solar farms with deep learning
 
Learning new climate science by thinking creatively with machine learning
Learning new climate science by thinking creatively with machine learningLearning new climate science by thinking creatively with machine learning
Learning new climate science by thinking creatively with machine learning
 
Neural Networks and Deep Learning for Physicists
Neural Networks and Deep Learning for PhysicistsNeural Networks and Deep Learning for Physicists
Neural Networks and Deep Learning for Physicists
 
Complex Models for Big Data
Complex Models for Big DataComplex Models for Big Data
Complex Models for Big Data
 
IBM Cloud Paris Meetup 20180517 - Deep Learning Challenges
IBM Cloud Paris Meetup 20180517 - Deep Learning ChallengesIBM Cloud Paris Meetup 20180517 - Deep Learning Challenges
IBM Cloud Paris Meetup 20180517 - Deep Learning Challenges
 
The effects of visual realism on search tasks in mixed reality simulations-IE...
The effects of visual realism on search tasks in mixed reality simulations-IE...The effects of visual realism on search tasks in mixed reality simulations-IE...
The effects of visual realism on search tasks in mixed reality simulations-IE...
 
Himansu sahoo resume-ds
Himansu sahoo resume-dsHimansu sahoo resume-ds
Himansu sahoo resume-ds
 
PPT s12-machine vision-s2
PPT s12-machine vision-s2PPT s12-machine vision-s2
PPT s12-machine vision-s2
 
Practical computer vision-- A problem-driven approach towards learning CV/ML/DL
Practical computer vision-- A problem-driven approach towards learning CV/ML/DLPractical computer vision-- A problem-driven approach towards learning CV/ML/DL
Practical computer vision-- A problem-driven approach towards learning CV/ML/DL
 
TraitCapture: NextGen phenomics tools for lab and field [ComBio2015]
TraitCapture: NextGen phenomics tools for lab and field [ComBio2015]TraitCapture: NextGen phenomics tools for lab and field [ComBio2015]
TraitCapture: NextGen phenomics tools for lab and field [ComBio2015]
 
Microstructural Analysis and Machine Learning
Microstructural Analysis and Machine LearningMicrostructural Analysis and Machine Learning
Microstructural Analysis and Machine Learning
 
TERN eMAST : Observations and terrestrial ecosystem models : Terrestrial Ecos...
TERN eMAST : Observations and terrestrial ecosystem models : Terrestrial Ecos...TERN eMAST : Observations and terrestrial ecosystem models : Terrestrial Ecos...
TERN eMAST : Observations and terrestrial ecosystem models : Terrestrial Ecos...
 
Machine Learning ebook.pdf
Machine Learning ebook.pdfMachine Learning ebook.pdf
Machine Learning ebook.pdf
 
1_5_AI_edx_ml_51intro_240204_104838machine learning lecture 1
1_5_AI_edx_ml_51intro_240204_104838machine learning lecture 11_5_AI_edx_ml_51intro_240204_104838machine learning lecture 1
1_5_AI_edx_ml_51intro_240204_104838machine learning lecture 1
 
Machine learning in scientific workflows
Machine learning in scientific workflowsMachine learning in scientific workflows
Machine learning in scientific workflows
 
Introduction to the Artificial Intelligence and Computer Vision revolution
Introduction to the Artificial Intelligence and Computer Vision revolutionIntroduction to the Artificial Intelligence and Computer Vision revolution
Introduction to the Artificial Intelligence and Computer Vision revolution
 
Kmeans plusplus
Kmeans plusplusKmeans plusplus
Kmeans plusplus
 

More from Andreas Kamilaris

Experiences from the use of CovTracer: A contact tracing tool deployed in Cyp...
Experiences from the use of CovTracer: A contact tracing tool deployed in Cyp...Experiences from the use of CovTracer: A contact tracing tool deployed in Cyp...
Experiences from the use of CovTracer: A contact tracing tool deployed in Cyp...Andreas Kamilaris
 
Transferring manure from livestock farms to be used as fertilizer in crop fields
Transferring manure from livestock farms to be used as fertilizer in crop fieldsTransferring manure from livestock farms to be used as fertilizer in crop fields
Transferring manure from livestock farms to be used as fertilizer in crop fieldsAndreas Kamilaris
 
Geospatial Analysis and Internet of Things in Environmental Informatics
Geospatial Analysis and Internet of Things in Environmental InformaticsGeospatial Analysis and Internet of Things in Environmental Informatics
Geospatial Analysis and Internet of Things in Environmental InformaticsAndreas Kamilaris
 
A Review on the Application of Natural Computing in Environmental Informatics
A Review on the Application of Natural Computing in Environmental InformaticsA Review on the Application of Natural Computing in Environmental Informatics
A Review on the Application of Natural Computing in Environmental InformaticsAndreas Kamilaris
 
The evolution of pervasive computing towards a Web of Things
The evolution of pervasive computing towards a Web of ThingsThe evolution of pervasive computing towards a Web of Things
The evolution of pervasive computing towards a Web of ThingsAndreas Kamilaris
 
AgriBigCAT: An Online Platform for Estimating the Impact of Livestock Agricul...
AgriBigCAT: An Online Platform for Estimating the Impact of Livestock Agricul...AgriBigCAT: An Online Platform for Estimating the Impact of Livestock Agricul...
AgriBigCAT: An Online Platform for Estimating the Impact of Livestock Agricul...Andreas Kamilaris
 
Estimating the Environmental Impact of Agriculture by means of Geospatial and...
Estimating the Environmental Impact of Agriculture by means of Geospatial and...Estimating the Environmental Impact of Agriculture by means of Geospatial and...
Estimating the Environmental Impact of Agriculture by means of Geospatial and...Andreas Kamilaris
 
Disaster Monitoring using Unmanned Aerial Vehicles and Deep Learning
Disaster Monitoring using Unmanned Aerial Vehicles and Deep LearningDisaster Monitoring using Unmanned Aerial Vehicles and Deep Learning
Disaster Monitoring using Unmanned Aerial Vehicles and Deep LearningAndreas Kamilaris
 
A Web of Things Based Eco-System for Urban Computing - Towards Smarter Cities
A Web of Things Based Eco-System for Urban Computing - Towards Smarter CitiesA Web of Things Based Eco-System for Urban Computing - Towards Smarter Cities
A Web of Things Based Eco-System for Urban Computing - Towards Smarter CitiesAndreas Kamilaris
 
Big data analysis and Integration of Geophysical information from the Catalan...
Big data analysis and Integration of Geophysical information from the Catalan...Big data analysis and Integration of Geophysical information from the Catalan...
Big data analysis and Integration of Geophysical information from the Catalan...Andreas Kamilaris
 
Estimating the Impact of Agriculture on the Environment of Catalunya by means...
Estimating the Impact of Agriculture on the Environment of Catalunya by means...Estimating the Impact of Agriculture on the Environment of Catalunya by means...
Estimating the Impact of Agriculture on the Environment of Catalunya by means...Andreas Kamilaris
 
Agri-IoT: A Semantic Framework for Internet of Things-enabled Smart Farming A...
Agri-IoT: A Semantic Framework for Internet of Things-enabled Smart Farming A...Agri-IoT: A Semantic Framework for Internet of Things-enabled Smart Farming A...
Agri-IoT: A Semantic Framework for Internet of Things-enabled Smart Farming A...Andreas Kamilaris
 
Do ”Web of Things Platforms” Truly Follow the Web of Things?
Do ”Web of Things Platforms” Truly Follow the Web of Things?Do ”Web of Things Platforms” Truly Follow the Web of Things?
Do ”Web of Things Platforms” Truly Follow the Web of Things?Andreas Kamilaris
 
WOTS2E: A Search Engine for a Semantic Web of Things
WOTS2E: A Search Engine for a Semantic Web of ThingsWOTS2E: A Search Engine for a Semantic Web of Things
WOTS2E: A Search Engine for a Semantic Web of ThingsAndreas Kamilaris
 
Enabling the physical world to the Internet and potential benefits for agricu...
Enabling the physical world to the Internet and potential benefits for agricu...Enabling the physical world to the Internet and potential benefits for agricu...
Enabling the physical world to the Internet and potential benefits for agricu...Andreas Kamilaris
 
Privacy Concerns in Sharing Personal Consumption Data through Online Applicat...
Privacy Concerns in Sharing Personal Consumption Data through Online Applicat...Privacy Concerns in Sharing Personal Consumption Data through Online Applicat...
Privacy Concerns in Sharing Personal Consumption Data through Online Applicat...Andreas Kamilaris
 
Social Electricity User Manual
Social Electricity User ManualSocial Electricity User Manual
Social Electricity User ManualAndreas Kamilaris
 
Social Electricity Online Platform (SEOP) EU Project Description
Social Electricity Online Platform (SEOP) EU Project DescriptionSocial Electricity Online Platform (SEOP) EU Project Description
Social Electricity Online Platform (SEOP) EU Project DescriptionAndreas Kamilaris
 
How the Internet can motivate you to switch off the lights
How the Internet can motivate you to switch off the lightsHow the Internet can motivate you to switch off the lights
How the Internet can motivate you to switch off the lightsAndreas Kamilaris
 

More from Andreas Kamilaris (20)

Experiences from the use of CovTracer: A contact tracing tool deployed in Cyp...
Experiences from the use of CovTracer: A contact tracing tool deployed in Cyp...Experiences from the use of CovTracer: A contact tracing tool deployed in Cyp...
Experiences from the use of CovTracer: A contact tracing tool deployed in Cyp...
 
Transferring manure from livestock farms to be used as fertilizer in crop fields
Transferring manure from livestock farms to be used as fertilizer in crop fieldsTransferring manure from livestock farms to be used as fertilizer in crop fields
Transferring manure from livestock farms to be used as fertilizer in crop fields
 
Geospatial Analysis and Internet of Things in Environmental Informatics
Geospatial Analysis and Internet of Things in Environmental InformaticsGeospatial Analysis and Internet of Things in Environmental Informatics
Geospatial Analysis and Internet of Things in Environmental Informatics
 
A Review on the Application of Natural Computing in Environmental Informatics
A Review on the Application of Natural Computing in Environmental InformaticsA Review on the Application of Natural Computing in Environmental Informatics
A Review on the Application of Natural Computing in Environmental Informatics
 
The evolution of pervasive computing towards a Web of Things
The evolution of pervasive computing towards a Web of ThingsThe evolution of pervasive computing towards a Web of Things
The evolution of pervasive computing towards a Web of Things
 
AgriBigCAT: An Online Platform for Estimating the Impact of Livestock Agricul...
AgriBigCAT: An Online Platform for Estimating the Impact of Livestock Agricul...AgriBigCAT: An Online Platform for Estimating the Impact of Livestock Agricul...
AgriBigCAT: An Online Platform for Estimating the Impact of Livestock Agricul...
 
Estimating the Environmental Impact of Agriculture by means of Geospatial and...
Estimating the Environmental Impact of Agriculture by means of Geospatial and...Estimating the Environmental Impact of Agriculture by means of Geospatial and...
Estimating the Environmental Impact of Agriculture by means of Geospatial and...
 
Disaster Monitoring using Unmanned Aerial Vehicles and Deep Learning
Disaster Monitoring using Unmanned Aerial Vehicles and Deep LearningDisaster Monitoring using Unmanned Aerial Vehicles and Deep Learning
Disaster Monitoring using Unmanned Aerial Vehicles and Deep Learning
 
A Web of Things Based Eco-System for Urban Computing - Towards Smarter Cities
A Web of Things Based Eco-System for Urban Computing - Towards Smarter CitiesA Web of Things Based Eco-System for Urban Computing - Towards Smarter Cities
A Web of Things Based Eco-System for Urban Computing - Towards Smarter Cities
 
Big data analysis and Integration of Geophysical information from the Catalan...
Big data analysis and Integration of Geophysical information from the Catalan...Big data analysis and Integration of Geophysical information from the Catalan...
Big data analysis and Integration of Geophysical information from the Catalan...
 
Estimating the Impact of Agriculture on the Environment of Catalunya by means...
Estimating the Impact of Agriculture on the Environment of Catalunya by means...Estimating the Impact of Agriculture on the Environment of Catalunya by means...
Estimating the Impact of Agriculture on the Environment of Catalunya by means...
 
Agri-IoT: A Semantic Framework for Internet of Things-enabled Smart Farming A...
Agri-IoT: A Semantic Framework for Internet of Things-enabled Smart Farming A...Agri-IoT: A Semantic Framework for Internet of Things-enabled Smart Farming A...
Agri-IoT: A Semantic Framework for Internet of Things-enabled Smart Farming A...
 
Do ”Web of Things Platforms” Truly Follow the Web of Things?
Do ”Web of Things Platforms” Truly Follow the Web of Things?Do ”Web of Things Platforms” Truly Follow the Web of Things?
Do ”Web of Things Platforms” Truly Follow the Web of Things?
 
WOTS2E: A Search Engine for a Semantic Web of Things
WOTS2E: A Search Engine for a Semantic Web of ThingsWOTS2E: A Search Engine for a Semantic Web of Things
WOTS2E: A Search Engine for a Semantic Web of Things
 
Enabling the physical world to the Internet and potential benefits for agricu...
Enabling the physical world to the Internet and potential benefits for agricu...Enabling the physical world to the Internet and potential benefits for agricu...
Enabling the physical world to the Internet and potential benefits for agricu...
 
Privacy Concerns in Sharing Personal Consumption Data through Online Applicat...
Privacy Concerns in Sharing Personal Consumption Data through Online Applicat...Privacy Concerns in Sharing Personal Consumption Data through Online Applicat...
Privacy Concerns in Sharing Personal Consumption Data through Online Applicat...
 
Social Electricity User Manual
Social Electricity User ManualSocial Electricity User Manual
Social Electricity User Manual
 
Social Electricity
Social ElectricitySocial Electricity
Social Electricity
 
Social Electricity Online Platform (SEOP) EU Project Description
Social Electricity Online Platform (SEOP) EU Project DescriptionSocial Electricity Online Platform (SEOP) EU Project Description
Social Electricity Online Platform (SEOP) EU Project Description
 
How the Internet can motivate you to switch off the lights
How the Internet can motivate you to switch off the lightsHow the Internet can motivate you to switch off the lights
How the Internet can motivate you to switch off the lights
 

Recently uploaded

Advanced Machine Learning for Business Professionals
Advanced Machine Learning for Business ProfessionalsAdvanced Machine Learning for Business Professionals
Advanced Machine Learning for Business ProfessionalsVICTOR MAESTRE RAMIREZ
 
modul pembelajaran robotic Workshop _ by Slidesgo.pptx
modul pembelajaran robotic Workshop _ by Slidesgo.pptxmodul pembelajaran robotic Workshop _ by Slidesgo.pptx
modul pembelajaran robotic Workshop _ by Slidesgo.pptxaleedritatuxx
 
RS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝Delhi
RS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝DelhiRS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝Delhi
RS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝Delhijennyeacort
 
Predicting Salary Using Data Science: A Comprehensive Analysis.pdf
Predicting Salary Using Data Science: A Comprehensive Analysis.pdfPredicting Salary Using Data Science: A Comprehensive Analysis.pdf
Predicting Salary Using Data Science: A Comprehensive Analysis.pdfBoston Institute of Analytics
 
INTERNSHIP ON PURBASHA COMPOSITE TEX LTD
INTERNSHIP ON PURBASHA COMPOSITE TEX LTDINTERNSHIP ON PURBASHA COMPOSITE TEX LTD
INTERNSHIP ON PURBASHA COMPOSITE TEX LTDRafezzaman
 
Defining Constituents, Data Vizzes and Telling a Data Story
Defining Constituents, Data Vizzes and Telling a Data StoryDefining Constituents, Data Vizzes and Telling a Data Story
Defining Constituents, Data Vizzes and Telling a Data StoryJeremy Anderson
 
Generative AI for Social Good at Open Data Science East 2024
Generative AI for Social Good at Open Data Science East 2024Generative AI for Social Good at Open Data Science East 2024
Generative AI for Social Good at Open Data Science East 2024Colleen Farrelly
 
Student Profile Sample report on improving academic performance by uniting gr...
Student Profile Sample report on improving academic performance by uniting gr...Student Profile Sample report on improving academic performance by uniting gr...
Student Profile Sample report on improving academic performance by uniting gr...Seán Kennedy
 
RadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdfRadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdfgstagge
 
Real-Time AI Streaming - AI Max Princeton
Real-Time AI  Streaming - AI Max PrincetonReal-Time AI  Streaming - AI Max Princeton
Real-Time AI Streaming - AI Max PrincetonTimothy Spann
 
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...Boston Institute of Analytics
 
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024thyngster
 
ASML's Taxonomy Adventure by Daniel Canter
ASML's Taxonomy Adventure by Daniel CanterASML's Taxonomy Adventure by Daniel Canter
ASML's Taxonomy Adventure by Daniel Cantervoginip
 
Thiophen Mechanism khhjjjjjjjhhhhhhhhhhh
Thiophen Mechanism khhjjjjjjjhhhhhhhhhhhThiophen Mechanism khhjjjjjjjhhhhhhhhhhh
Thiophen Mechanism khhjjjjjjjhhhhhhhhhhhYasamin16
 
Minimizing AI Hallucinations/Confabulations and the Path towards AGI with Exa...
Minimizing AI Hallucinations/Confabulations and the Path towards AGI with Exa...Minimizing AI Hallucinations/Confabulations and the Path towards AGI with Exa...
Minimizing AI Hallucinations/Confabulations and the Path towards AGI with Exa...Thomas Poetter
 
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort servicejennyeacort
 
办美国阿肯色大学小石城分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degree
办美国阿肯色大学小石城分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degree办美国阿肯色大学小石城分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degree
办美国阿肯色大学小石城分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degreeyuu sss
 
Conf42-LLM_Adding Generative AI to Real-Time Streaming Pipelines
Conf42-LLM_Adding Generative AI to Real-Time Streaming PipelinesConf42-LLM_Adding Generative AI to Real-Time Streaming Pipelines
Conf42-LLM_Adding Generative AI to Real-Time Streaming PipelinesTimothy Spann
 
Statistics, Data Analysis, and Decision Modeling, 5th edition by James R. Eva...
Statistics, Data Analysis, and Decision Modeling, 5th edition by James R. Eva...Statistics, Data Analysis, and Decision Modeling, 5th edition by James R. Eva...
Statistics, Data Analysis, and Decision Modeling, 5th edition by James R. Eva...ssuserf63bd7
 
Learn How Data Science Changes Our World
Learn How Data Science Changes Our WorldLearn How Data Science Changes Our World
Learn How Data Science Changes Our WorldEduminds Learning
 

Recently uploaded (20)

Advanced Machine Learning for Business Professionals
Advanced Machine Learning for Business ProfessionalsAdvanced Machine Learning for Business Professionals
Advanced Machine Learning for Business Professionals
 
modul pembelajaran robotic Workshop _ by Slidesgo.pptx
modul pembelajaran robotic Workshop _ by Slidesgo.pptxmodul pembelajaran robotic Workshop _ by Slidesgo.pptx
modul pembelajaran robotic Workshop _ by Slidesgo.pptx
 
RS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝Delhi
RS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝DelhiRS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝Delhi
RS 9000 Call In girls Dwarka Mor (DELHI)⇛9711147426🔝Delhi
 
Predicting Salary Using Data Science: A Comprehensive Analysis.pdf
Predicting Salary Using Data Science: A Comprehensive Analysis.pdfPredicting Salary Using Data Science: A Comprehensive Analysis.pdf
Predicting Salary Using Data Science: A Comprehensive Analysis.pdf
 
INTERNSHIP ON PURBASHA COMPOSITE TEX LTD
INTERNSHIP ON PURBASHA COMPOSITE TEX LTDINTERNSHIP ON PURBASHA COMPOSITE TEX LTD
INTERNSHIP ON PURBASHA COMPOSITE TEX LTD
 
Defining Constituents, Data Vizzes and Telling a Data Story
Defining Constituents, Data Vizzes and Telling a Data StoryDefining Constituents, Data Vizzes and Telling a Data Story
Defining Constituents, Data Vizzes and Telling a Data Story
 
Generative AI for Social Good at Open Data Science East 2024
Generative AI for Social Good at Open Data Science East 2024Generative AI for Social Good at Open Data Science East 2024
Generative AI for Social Good at Open Data Science East 2024
 
Student Profile Sample report on improving academic performance by uniting gr...
Student Profile Sample report on improving academic performance by uniting gr...Student Profile Sample report on improving academic performance by uniting gr...
Student Profile Sample report on improving academic performance by uniting gr...
 
RadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdfRadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdf
 
Real-Time AI Streaming - AI Max Princeton
Real-Time AI  Streaming - AI Max PrincetonReal-Time AI  Streaming - AI Max Princeton
Real-Time AI Streaming - AI Max Princeton
 
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
 
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
 
ASML's Taxonomy Adventure by Daniel Canter
ASML's Taxonomy Adventure by Daniel CanterASML's Taxonomy Adventure by Daniel Canter
ASML's Taxonomy Adventure by Daniel Canter
 
Thiophen Mechanism khhjjjjjjjhhhhhhhhhhh
Thiophen Mechanism khhjjjjjjjhhhhhhhhhhhThiophen Mechanism khhjjjjjjjhhhhhhhhhhh
Thiophen Mechanism khhjjjjjjjhhhhhhhhhhh
 
Minimizing AI Hallucinations/Confabulations and the Path towards AGI with Exa...
Minimizing AI Hallucinations/Confabulations and the Path towards AGI with Exa...Minimizing AI Hallucinations/Confabulations and the Path towards AGI with Exa...
Minimizing AI Hallucinations/Confabulations and the Path towards AGI with Exa...
 
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
 
办美国阿肯色大学小石城分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degree
办美国阿肯色大学小石城分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degree办美国阿肯色大学小石城分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degree
办美国阿肯色大学小石城分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degree
 
Conf42-LLM_Adding Generative AI to Real-Time Streaming Pipelines
Conf42-LLM_Adding Generative AI to Real-Time Streaming PipelinesConf42-LLM_Adding Generative AI to Real-Time Streaming Pipelines
Conf42-LLM_Adding Generative AI to Real-Time Streaming Pipelines
 
Statistics, Data Analysis, and Decision Modeling, 5th edition by James R. Eva...
Statistics, Data Analysis, and Decision Modeling, 5th edition by James R. Eva...Statistics, Data Analysis, and Decision Modeling, 5th edition by James R. Eva...
Statistics, Data Analysis, and Decision Modeling, 5th edition by James R. Eva...
 
Learn How Data Science Changes Our World
Learn How Data Science Changes Our WorldLearn How Data Science Changes Our World
Learn How Data Science Changes Our World
 

Training deep learning models to count using synthetic images

  • 1. TRAINING DEEP LEARNING MODELS TO COUNT USING SYNTHETIC IMAGES 6 Sept. 2019DR. ANDREAS KAMILARIS @DL-UAV19, PROC. OF CAIP 2019
  • 2. Problem • Difficult to create ground truth data • UCSD Pedestrian Database • Video of pedestrians on UCSD walkways, taken from a stationary camera
  • 3. Problem • Difficult to create ground truth data • UCF CC 50 dataset • Counts of persons range between 94 and 4543, with an average of 1280 individuals per image
  • 4. Motivation • Generate data by simulation, which are easy then to label. • Create synthetic ground truth data Rahnemoonfar, M. and Sheppard, C., 2017. Deep count: fruit counting based on deep simulated learning. Sensors, 17(4), p.905.
  • 5. First try • Generated simulated data in Python • Goal: Detection of fires in forests • 160 images, 80 of forest, 80 of fire • 80% training, 20% testing
  • 6. First try • Model: Inception-v3 vs. Custom
  • 7. First try • False negatives where there was smoke AND fire! Data Custom Inception-v3 Only generative 0.65 0.62 Only real 0.8 0.69 Combined, augmented dataset 0.9 0.71
  • 8. Second try • Better representation of fire
  • 9. Second try • Improvement of results! • Very few false negatives Data Custom Inception-v3 Only generative 0.95 0.72 Only real 0.8 0.69 Combined, augmented dataset 1.0 0.79
  • 10. Generative data case Training based on 2,000 synthetic images (labelled as fire or forest) Testing based on 100 real-world aerial photos (classified as 50 images of forest and 50 images of fire)
  • 11. Hypothesis 1 Generating synthetic data can help to train deep learning models, without the need to create expensive (in terms of time and effort) ground truth data!
  • 12. Hypothesis 2 Generating synthetic data can help to train deep learning models not only to classify, but also to count! … not only simple problems, but also more advanced ones…
  • 13. Application: Counting houses from aerial photos  60 photos taken from satellite images in urban areas of Tanzania  Manually counted the number of houses, to create out testing/validation dataset.  Each photo has [0,38] houses.
  • 14. Application: Counting houses from aerial photos  Created synthetic training data, with automated house counting  First naïve try: Involved only squares! MSE = 41
  • 15. Application: Counting houses from aerial photos  Second try: Added trees and small shadows MSE = 29
  • 16. Application: Counting houses from aerial photos  Third try: Added grass, fences and different orientation of houses. Also added images without any houses MSE = 20
  • 17. Application: Counting houses from aerial photos  Many decisions to take along the way… • Dropout rate (35-50% works well) • Stride (2 is small, 10 is too big) • Convolutions (7x7 initially seems a good option) • Pre-training (ImageNet is not helping a lot) • Max-pooling better than average pooling • Dense layers at the end of the network, ReLU function
  • 18. Application: Counting houses from aerial photos Adapted, custom topology Inception-ResNet 7x7 input filter with large stride Dense fully- connected layer with ReLU
  • 19. Application: Counting houses from aerial photos  Training vs. Testing MSE The model can predict the number of houses with an error of 4,47 houses. For example, for a photo with 20 houses, the model would predict in the range of [16, 24]. Training based on 10,000 synthetic images (labelled with exact number of houses) Testing based on 60 real- world aerial photos (labelled with exact number of houses)
  • 20. Application: Counting houses from aerial photos  Best Vs. Worst predictions
  • 21. Application: Counting houses from aerial photos  Next steps: • Crop houses from training dataset and reuse based on random combinations in semi-synthetic images • More realistic generation of data • (GAN for counting? ) • Accountability • Other domains: o Agriculture (counting animals in farms) o Energy (renewable energy in roofs) o Environment and Climate (counting trees, plants, endangered species of animals etc.) o Microbiology (blood test analysis etc.)
  • 22.  State of the art (published in 2019) Kar, A., Prakash, A., Liu, M.Y., Cameracci, E., Yuan, J., Rusiniak, M., Acuna, D., Torralba, A. and Fidler, S., 2019. Meta-Sim: Learning to Generate Synthetic Datasets. arXiv preprint arXiv:1904.11621. Future Work and Research Direction
  • 23. Future Work and Research Direction  State of the art (published in 2019)  Combining Counting CNN model with the ResNeXt architecture Tian, M. et al., Automated pig counting using deep learning, Computers and Electronics in Agriculture, vol. 163, pp. 1-10, 2019. MAE = 4.47 MAE = 2.77
  • 24. Conclusion • Synthetic data can be used for training DL models • Can be applied in UAV-related applications (classification vs. counting problems) • More advanced techniques are required for improving performance (e.g. probabilistic scene graph generation, density maps)
  • 25. THANKS FOR YOUR ATTENTION! DR. ANDREAS KAMILARIS EMAIL: A.KAMILARIS@UTWENTE.NL