Processing images with Deep Learning

•

0 likes•127 views

Amazon Rekognition or Apache MXNet? Github projects for image processing with Apache MXNet A deeper look at the Convolution operation Demos

Technology

Processing images
with Deep Learning
Julien Simon, AI Evangelist, EMEA
@julsimon

What to expect
• Amazon Rekognition or Apache MXNet?
• Github projects for image processing with Apache MXNet
• A deeper look at the Convolution operation
• Demos
• Q&A

Apache MXNet: Open Source library for Deep Learning
Programmable Portable High Performance
Near linear scaling
across hundreds of
GPUs
Highly efficient
models for
mobile
and IoT
Simple syntax,
multiple
languages
Most Open Best On AWS
Optimized for
Deep Learning on AWS
Accepted into the
Apache Incubator

Object Detection
https://github.com/precedenceguo/mx-rcnn https://github.com/zhreshold/mxnet-yolo

Object Segmentation
https://github.com/TuSimple/mx-maskrcnn

Text Detection and Recognition
https://github.com/Bartzi/stn-ocr

Real-Time Pose Estimation
https://github.com/dragonfly90/mxnet_Realtime_Multi-Person_Pose_Estimation

Demos
https://github.com/juliensimon/dlnotebooks
https://github.com/guyernest/TensorFlowTutorials
1) Classifying MNIST with a CNN model (Keras)
2) Classifying images with pre-trained CNN models (MXNet)
3) Fine-tuning a pre-trained CNN model (Keras)
4) Generating new MNIST samples with a GAN (MXNet)

Demo #2 – Using a pre-trained model
*** VGG16
[(0.46811387, 'n04296562 stage'), (0.24333163,
'n03272010 electric guitar'), (0.045918692, 'n02231487
walking stick, walkingstick, stick insect'),
(0.03316205, 'n04286575 spotlight, spot'),
(0.021694135, 'n03691459 loudspeaker, speaker, speaker
unit, loudspeaker system, speaker system')]
*** ResNet-152
[(0.8726753, 'n04296562 stage'), (0.046159592,
'n03272010 electric guitar'), (0.041658506, 'n03759954
microphone, mike'), (0.018624334, 'n04286575 spotlight,
spot'), (0.0058045341, 'n02676566 acoustic guitar')]
*** Inception v3
[(0.44991142, 'n04296562 stage'), (0.43065304,
'n03272010 electric guitar'), (0.067580454, 'n04456115
torch'), (0.012423956, 'n02676566 acoustic guitar'),
(0.0093934005, 'n03250847 drumstick')]
https://medium.com/@julsimon/an-introduction-to-the-mxnet-api-part-5-9e78534096db

Demo #3 – Image classification: fine-tuning a model
• CIFAR-10 data set
• 60,000 images in 10 classes
• 32x32 color images
• Initial training
• Resnet-50 CNN
• 200 epochs
• 82.12% validation
• Cars vs. horses
• 88.8% validation accuracy
https://medium.com/@julsimon/keras-shoot-out-part-3-fine-tuning-7d1548c51a41

Demo #3 – Image classification: fine-tuning a model
• Freezing all layers but the last one
• Fine-tuning on « cars vs. horses » for 10 epochs
• 2 minutes on 1 GPU
• 98.8% validation accuracy
Epoch 10/10
10000/10000 [==============================] - 12s
loss: 1.6989 - acc: 0.9994 - val_loss: 1.7490 - val_acc: 0.9880
2000/2000 [==============================] - 2s
[1.7490020694732666, 0.98799999999999999]

© 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Resources
https://aws.amazon.com/machine-learning
https://aws.amazon.com/blogs/ai
https://mxnet.incubator.apache.org
https://github.com/apache/incubator-mxnet
https://github.com/gluon-api
https://devblogs.nvidia.com/parallelforall/deep-learning-nutshell-core-concepts/
http://deeplearning.net/software/theano/tutorial/conv_arithmetic.html
https://medium.com/@julsimon

Thank you!
Julien Simon, AI Evangelist, EMEA
@julsimon

How do you think about computing resources in a world where you can launch and terminate computational capacity in minutes? Amazon EC2 provides a powerful platform to access vast computational resources at the click of a button or a simple API call. It is also very different from operating your own data center or having to managed fix assets in a co-location facility. This talk walks you through examples of how the cloud enables more efficient capacity planning, provides guidance in how developers and organizations can manage thousands of instances efficiently, and highlights tools that make it easy for you to plan your capacity needs, even when those needs might require you to provision the equivalent of a small data center at short notice.

Deep learning an Introduction with Competitive Landscape

Shivaji Dutta

This gives an introduction to Neural Networks to CNN, RNN, Reinforcement Learning to what competitive tools are out there. Also a comparison of the various frameworks from Tensorflow, Caffe, Chainer and Pytorch. We also capture the work done by various other companies in the enterprise tools space, web service offerings from Google, Sales Force and Amazon. End we mention the various Heroes of the Deep Learning space.

Deep Dive on Deep Learning with Apache MXNet - AWS Summit Tel Aviv 2017

Amazon Web Services

Deep learning continues to push the state of the art in domains such as computer vision, natural language understanding, and recommendation engines. One of the key reasons for this progress is the availability of highly flexible, and developer friendly deep learning frameworks. Apache MXNet is a fully-featured, flexible, and massively scalable open source deep learning framework that supports innovative deep models, including convolutional neural networks (CNNs) and long short-term memory networks (LSTMs). This session will show you how to use the AWS Deep Learning AMI and Cloud Formation template to deploy and train your own deep neural network, using MNIST, to recognize handwritten digits and test it for accuracy as a practical introduction to Apache MXNet.

Deep Learning on Qubole Data Platform

Shivaji Dutta

Systems Bioinformatics Workshop KeynoteDeepak Singh

Azure Batch AI for Neural Networks

Cameron Vetter

Do you want to predict customer behavior? Evaluate the content of a photo or sound? Detect Fraud? Feed usage data back into your algorithms to improve them automatically? All of these things are being done today using Neural Networks for Machine Learning. This talk will cover how to use the GPU power of Azure to train a Neural Network and how to turn that Neural Network into a REST service hosted in Azure. The topics covered include: • Brief overview of Neural Networks • Azure Batch AI • Azure Data Science Virtual Machines • Python in Azure Web Apps You'll leave with an understanding of how to use Azure to train and host your neural networks.

Distributed deep learning optimizations - AI WithTheBest

geetachauhan

Metta Innovations - Introdução ao Deep Learning aplicado a vídeo analytics

Eduardo Gaspar

Applying your Convolutional Neural Networks

Databricks

Part 3 of the Deep Learning Fundamentals Series, this session starts with a quick primer on activation functions, learning rates, optimizers, and backpropagation. Then it dives deeper into convolutional neural networks discussing convolutions (including kernels, local connectivity, strides, padding, and activation functions), pooling (or subsampling to reduce the image size), and fully connected layer. The session also provides a high-level overview of some CNN architectures. The demos included in these slides are running on Keras with TensorFlow backend on Databricks.

Deep Learning Update May 2016

Frédéric Parienté

The Revolution of Deep Learning

Frédéric Parienté

Deep Learning

Büşra İçöz

Android and Deep Learning

Oswald Campesato

Deep Learning, Microsoft Cognitive Toolkit (CNTK) and Azure Machine Learning ...

Naoki (Neo) SATO

An Introduction to Deep Learning with Apache MXNet (November 2017)

Julien SIMON

運用CNTK 實作深度學習物件辨識 Deep Learning based Object Detection with Microsoft Cogniti...

Herman Wu

What's hot

Deep Learning for Developers (October 2017)

Julien SIMON

Deep Learning for Developers (October 2017)

Julien SIMON

Deep Learning for Developers (December 2017)

Julien SIMON

MXNet Workshop

Amazon Web Services

Deep Learning Computer Build

PetteriTeikariPhD

An introduction to Deep Learning with Apache MXNet (November 2017)

Julien SIMON

CPN211 My Datacenter Has Walls That Move - AWS re: Invent 2012

Amazon Web Services

Deep learning an Introduction with Competitive Landscape

Shivaji Dutta

Deep Dive on Deep Learning with Apache MXNet - AWS Summit Tel Aviv 2017

Amazon Web Services

Deep Learning on Qubole Data Platform

Shivaji Dutta

Systems Bioinformatics Workshop KeynoteDeepak Singh

Azure Batch AI for Neural Networks

Cameron Vetter

Distributed deep learning optimizations - AI WithTheBest

geetachauhan

Metta Innovations - Introdução ao Deep Learning aplicado a vídeo analytics

Eduardo Gaspar

Applying your Convolutional Neural Networks

Databricks

Deep Learning Update May 2016

Frédéric Parienté

The Revolution of Deep Learning

Frédéric Parienté

Deep Learning

Büşra İçöz

Android and Deep Learning

Oswald Campesato

Deep Learning, Microsoft Cognitive Toolkit (CNTK) and Azure Machine Learning ...

Naoki (Neo) SATO

What's hot (20)

Deep Learning for Developers (October 2017)

Deep Learning for Developers (December 2017)

MXNet Workshop

Deep Learning Computer Build

An introduction to Deep Learning with Apache MXNet (November 2017)

CPN211 My Datacenter Has Walls That Move - AWS re: Invent 2012

Deep learning an Introduction with Competitive Landscape

Deep Dive on Deep Learning with Apache MXNet - AWS Summit Tel Aviv 2017

Deep Learning on Qubole Data Platform

Systems Bioinformatics Workshop Keynote

Azure Batch AI for Neural Networks

Distributed deep learning optimizations - AI WithTheBest

Metta Innovations - Introdução ao Deep Learning aplicado a vídeo analytics

Applying your Convolutional Neural Networks

Deep Learning Update May 2016

The Revolution of Deep Learning

Deep Learning

Android and Deep Learning

Deep Learning, Microsoft Cognitive Toolkit (CNTK) and Azure Machine Learning ...

Similar to Processing images with Deep Learning

An Introduction to Deep Learning with Apache MXNet (November 2017)

Julien SIMON

運用CNTK 實作深度學習物件辨識 Deep Learning based Object Detection with Microsoft Cogniti...

Herman Wu

Depth Imageからの Keypoint Detection

Tetsuro Kato

Introduction to Neural Networks

Databricks

Part 1 of the Deep Learning Fundamentals Series, this session discusses the use cases and scenarios surrounding Deep Learning and AI; reviews the fundamentals of artificial neural networks (ANNs) and perceptrons; discuss the basics around optimization beginning with the cost function, gradient descent, and backpropagation; and activation functions (including Sigmoid, TanH, and ReLU). The demos included in these slides are running on Keras with TensorFlow backend on Databricks.

Panoramic Video in Environmental Monitoring Software Development and Applica...

pycontw

Long Term Recurrent Convolutional Neural Networks

Ekin Akyürek

Raspberry Pi: Python todo en uno para dummies por John Shovic parte 2.pdf

SANTIAGO PABLO ALBERTO

Solr and Machine Vision - Scott Cote, Lucidworks & Trevor Grant, IBM

Lucidworks

TestowanieIoT2016

kraqa

Amazon SageMaker을 통한 손쉬운 Jupyter Notebook 활용하기 - 윤석찬 (AWS 테크에반젤리스트)

Amazon Web Services Korea

Amazon SageMaker는 기계 학습을 위한 데이터와 알고리즘, 프레임워크를 빠르게 연결하에 손쉽게 ML 구축이 가능한 신규 클라우드 서비스입니다. 이번 시간에는 Amazon S3에 저장된 학습 데이터를 이용하여 가장 일반적으로 사용하는 알고리즘 몇 가지를 직접 실행해 보는 실습을 진행합니다. 이를 위해 유명한 오픈 소스 프레임워크인 TensorFlow와 Keras 그리고 Apache MXNet과 Gluon 등을 사용해 봅니다.

Convolutional Neural Networks for Computer vision Applications

Alex Conway

DevExperience 2018 : Building a Smart Security Camera with Raspberry Pi Zero,...

Mark West

In this session, I’ll share how I transformed a lowly Raspberry Pi Zero W webcam into a smart security camera (with motion detection, threat analysis and alert notifications) by combining open source software with cloud based image analysis. Attendees can expect a short explanation of how to set up their own motion activated webcam and a demonstration of how they can use Java and a range of AWS Services (including Rekognition, Lambda Functions and Step Functions) to help their camera distinguish between an unwanted guest and the neighbour’s cat.

Deep Learning with Audio Signals: Prepare, Process, Design, Expect

Keunwoo Choi

Is deep learning Alchemy? No! But it heavily relies on tips and tricks, a set of common wisdom that probably works for similar problems. In this talk, I’ll introduce what the audio/music research societies have discovered while playing with deep learning when it comes to audio classification and regression -- how to prepare the audio data and preprocess them, how to design the networks (or choose which one to steal from), and what we can expect as a result.

Meetup web scale architecture quantum computing (Part 1 16-10-2018)

Rolf Huisman

Sumatra and git

Michele Mattioni

Machine Learning Tokyo - Deep Neural Networks for Video - NumberBoost

Alex Conway

Separating Hype from Reality in Deep Learning with Sameer Farooqui

Databricks

Deep Learning is all the rage these days, but where does the reality of what Deep Learning can do end and the media hype begin? In this talk, I will dispel common myths about Deep Learning that are not necessarily true and help you decide whether you should practically use Deep Learning in your software stack. I’ll begin with a technical overview of common neural network architectures like CNNs, RNNs, GANs and their common use cases like computer vision, language understanding or unsupervised machine learning. Then I’ll separate the hype from reality around questions like: • When should you prefer traditional ML systems like scikit learn or Spark.ML instead of Deep Learning? • Do you no longer need to do careful feature extraction and standardization if using Deep Learning? • Do you really need terabytes of data when training neural networks or can you ‘steal’ pre-trained lower layers from public models by using transfer learning? • How do you decide which activation function (like ReLU, leaky ReLU, ELU, etc) or optimizer (like Momentum, AdaGrad, RMSProp, Adam, etc) to use in your neural network? • Should you randomly initialize the weights in your network or use more advanced strategies like Xavier or He initialization? • How easy is it to overfit/overtrain a neural network and what are the common techniques to ovoid overfitting (like l1/l2 regularization, dropout and early stopping)?

Information from pixels

Dave Snowdon

With cheap cameras becoming ubiquitous the camera has become probably the most important sensor for many applications. However extracting usable information from the images produced by cameras is non-trivial. There have been many published successes in recent years using deep learning (multi-layered convolutional neural networks) but it’s not always necessary to apply such techniques to get useful results for many applications. This talk will focus on “classical” machine vision using java and the OpenCV library. We’ll start with a quick refresher on how image data is represented and then cover topics such as determining if an image is blurred (and therefore unusable) and then explore a number of techniques such as shape and face detection.

Training Neural Networks

Databricks

Part 2 of the Deep Learning Fundamentals Series, this session discusses Tuning Training (including hyperparameters, overfitting/underfitting), Training Algorithms (including different learning rates, backpropagation), Optimization (including stochastic gradient descent, momentum, Nesterov Accelerated Gradient, RMSprop, Adaptive algorithms - Adam, Adadelta, etc.), and a primer on Convolutional Neural Networks. The demos included in these slides are running on Keras with TensorFlow backend on Databricks.

Deep Learning on iOS #360iDev

Shuichi Tsutsumi

Similar to Processing images with Deep Learning (20)

An Introduction to Deep Learning with Apache MXNet (November 2017)

運用CNTK 實作深度學習物件辨識 Deep Learning based Object Detection with Microsoft Cogniti...

Depth Imageからの Keypoint Detection

Introduction to Neural Networks

Panoramic Video in Environmental Monitoring Software Development and Applica...

Long Term Recurrent Convolutional Neural Networks

Raspberry Pi: Python todo en uno para dummies por John Shovic parte 2.pdf

Solr and Machine Vision - Scott Cote, Lucidworks & Trevor Grant, IBM

TestowanieIoT2016

Amazon SageMaker을 통한 손쉬운 Jupyter Notebook 활용하기 - 윤석찬 (AWS 테크에반젤리스트)

Convolutional Neural Networks for Computer vision Applications

DevExperience 2018 : Building a Smart Security Camera with Raspberry Pi Zero,...

Deep Learning with Audio Signals: Prepare, Process, Design, Expect

Meetup web scale architecture quantum computing (Part 1 16-10-2018)

Sumatra and git

Machine Learning Tokyo - Deep Neural Networks for Video - NumberBoost

Separating Hype from Reality in Deep Learning with Sameer Farooqui

Information from pixels

Training Neural Networks

Deep Learning on iOS #360iDev

More from Julien SIMON

An introduction to computer vision with Hugging Face

Julien SIMON

Reinventing Deep Learning  with Hugging Face Transformers

Julien SIMON

Building NLP applications with Transformers

Julien SIMON

Building Machine Learning Models Automatically (June 2020)

Julien SIMON

Starting your AI/ML project right (May 2020)

Julien SIMON

Scale Machine Learning from zero to millions of users (April 2020)

Julien SIMON

An Introduction to Generative Adversarial Networks (April 2020)

Julien SIMON

AIM410R1 Deep learning applications with TensorFlow, featuring Fannie Mae (De...

Julien SIMON

AIM361 Optimizing machine learning models with Amazon SageMaker (December 2019)

Julien SIMON

AIM410R Deep Learning Applications with TensorFlow, featuring Mobileye (Decem...

Julien SIMON

A pragmatic introduction to natural language processing models (October 2019)

Julien SIMON

Building smart applications with AWS AI services (October 2019)

Julien SIMON

Build, train and deploy ML models with SageMaker (October 2019)

Julien SIMON

The Future of AI (September 2019)

Julien SIMON

Building Machine Learning Inference Pipelines at Scale (July 2019)

Julien SIMON

Talk at OSCON, Portland, 18/07/2019 Real-life Machine Learning applications require more than a single model. Data may need pre-processing: normalization, feature engineering, dimensionality reduction, etc. Predictions may need post-processing: filtering, sorting, combining, etc. Our goal: build scalable ML pipelines with open source (Spark, Scikit-learn, XGBoost) and managed services (Amazon EMR, AWS Glue, Amazon SageMaker)

Train and Deploy Machine Learning Workloads with AWS Container Services (July...

Julien SIMON

Optimize your Machine Learning Workloads on AWS (July 2019)

Julien SIMON

Deep Learning on Amazon Sagemaker (July 2019)

Julien SIMON

Automate your Amazon SageMaker Workflows (July 2019)

Julien SIMON

Build, train and deploy ML models with Amazon SageMaker (May 2019)

Julien SIMON

More from Julien SIMON (20)