Amazon SageMaker로 딥 러닝 기반 이미지 검색 서비스 만들기

Agenda • Amazon Rekognition • Amazon SageMaker • SageMaker Studio
Notebooks • SageMaker Training Job • SageMaker Endpoints

AnyCompany Dream is

Which Advertisement Is Better?

(B) may be Better, isn’t It?

Why (B) is Better? Climbing sunglasses Climbing shoes How about
climbing gloves?

How to Choose the Most Relevant Advertisement? • may be
easy to find, if a image has correct tags. • If so, how to tag millions of images in fast, cost-efficient way? • Machine Learning may save your life

Automatic Image Tagging Automatic Image Recognition

Amazon SageMaker Label data Aggregate & prepare data Store &
share features Auto ML Spark/R Detect bias Visualize in notebooks Pick algorithm Train models Tune parameters Debug & profile Deploy in production Manage & monitor CI/CD Human review Ground Truth Data Wrangler Feature store Autopilot Processing Clarify Studio Notebooks Built-in or Bring-your-own Experiments Spot Training Distributed Training Automatic Model Tuning Debugger Model Hosting Multi-model Endpoints Model Monitor Pipelines Augmented AI AMAZON SAGEMAKER EDGE MANAGER SAGEMAKER STUDIO IDE AMAZON SAGEMAKER JUMPSTART VISION SPEECH TEXT SEARCH CHATBOTS PERSONALIZATION FORECASTING FRAUD CONTACT CENTERS Deep Learning AMIs & Containers GPUs & CPUs Elastic Inference Trainium Inferentia FPGA AI SERVICES ML SERVICES FRAMEWORKS & INFRASTRUCTURE DeepGraphLibrary Amazon Rekognition Amazon Polly Amazon Transcribe +Medical Amazon Lex Amazon Personalize Amazon Forecast Amazon Comprehend +Medical Amazon Textract Amazon Kendra Amazon CodeGuru Amazon Fraud Detector Amazon Translate INDUSTRIAL AI CODE AND DEVOPS Amazon DevOps Guru Voice ID For Amazon Connect Contact Lens Amazon Monitron AWS Panorama + Appliance Amazon Lookout for Vision Amazon Lookout for Equipment Amazon HealthLake HEALTHCARE AI Amazon Lookout for Metrics ANOMOLY DETECTION Amazon Transcribe for Medical Amazon Comprehend for Medical 모든 개발자를 위한 다양한 인공 지능 도구 제공

Amazon Rekognition 기계 학습을 통한 이미지 및 비디오 분석 자동화

Amazon Rekognition Image and Video CELEBRITY RECOGNITION FACE COMPARE &
SEARCH PATHING FACE DETECTION AND ANALYSIS LIVE STREAM VIDEO TEXT CONTENT MODERATION CUSTOM LABELS OBJECT, SCENE, AND ACTIVITY VIDEO SEGMENTS PPE DETECTION

Architecture: Automatic Image Tagging Amazon OpenSearch Service Amazon Kinesis Data
Streams Amazon Rekognition Images AWS Lambda Amazon S3 bucket with images Amazon API Gateway https://tinyurl.com/kxydd2wn Automatic Image Recognition

Demo: Automatic Image Tagging https://github.com/aws-samples/aws-realtime-image-analysis

1 Year Later

Visual Search Similar results bonzai Provence lavender Cozy places Top
10 cozy apartments in Brooklyn

Building a Visual Search Application https://aws.amazon.com/blogs/machine-learning/building-a-visual-search-application-with- amazon-sagemaker-and-amazon-es/

As-Is: Automatic Image Tagging Amazon Kinesis Data Streams Client AWS
Lambda Amazon S3 bucket with images Amazon API Gateway Amazon OpenSearch Service Amazon Rekognition

To-Be: Visual Search Amazon Kinesis Data Streams Client AWS Lambda
Amazon S3 bucket with images Amazon API Gateway Amazon OpenSearch Service

Machine Learning is like this Model Training Prediction Output New
Data Labeled Data Cat

Typical Machine Learning Process Collect, prepare and label training data
Choose and optimize ML algorithm Train and tune ML models Set up and manage environments for training Deploy models in production Scale and manage the production environment 1 2 3

Set up and track experiment Choose model Debug, compare, and
evaluate experiments Monitor quality, detect drift, and retrain Share, review, and collaborate Machine Learning is iterative

End-to-End Machine Learning Platform Zero setup Flexible Model Training Pay
by the second $ Amazon SageMaker 손쉬운 기계 학습 모델 생성, 훈련 및 서비스 배포 완전 관리 서비스

Amazon SageMaker Studio 기계 학습 모델 개발 및 배포를 위한
최초의 완전 통합 개발 환경 (IDE) Collaboration at scale 코드 의존성 추적 없이 확장 가능한 노트북 공유 Easy experiment management 수천 개의 모델 실험을 구성, 추적 및 비교 Automatic model generation 코드 작성 없이 데이터를 가지고 자동 모델 생성 Higher quality ML models 오류 자동 디버깅 및 실시간 오류 경보 모델 모니터링 및 고품질 유지 Increased productivity 완전 자동화된 머신 러닝 워크플로 구축

최초의 완전 통합 개발 환경 (IDE)

Amazon SageMaker Studio 시작 화면

Notebook Model Models bucket SageMaker Endpoint Train Amazon SageMaker End
to End Training and Deployment

From Amazon SageMaker Notebooks training

Amazon SageMaker Training Specify Training Infrastructure

Amazon SageMaker Training Use Algorithm and Start training Execution Role
SageMaker Estimator

Amazon SageMaker Training How does training happen XGBoost validation(optional) test(optional
ECR S3 ML Instance ml.m4.xlarge xgboost linear-learner PCA DeepAR BlazingText Image classification … Object Detection Images S3 SageMaker Notebook SageMaker Training Job train Model

Model Deployment training

Model artifacts Amazon SageMaker Deployment Hosting Services Inference Image Training
Image Training Data Model artifacts Endpoint Amazon SageMaker Amazon S3 Amazon ECR Inference Image

Amazon SageMaker Deployment SageMaker Endpoints (Private API) Auto Scaling group
Availability Zone 1 Availability Zone 2 Availability Zone 3 Elastic Load Balancing Model Endpoint Client Deployment / Hosting Amazon SageMaker ML Compute Instances Input Data (Request) Prediction (Response)

Amazon SageMaker Deployment SageMaker Endpoints (Public API) Auto Scaling group
Availability Zone 1 Availability Zone 2 Availability Zone 3 Elastic Load Balancing Model Endpoint Amazon API Gateway Client Deployment / Hosting Amazon SageMaker ML Compute Instances Input Data (Request) Prediction (Response)

Define Estimator Object created deploy() predict() Object created fit() Amazon
SageMaker End to End Training and Deployment

Classification • Linear Learner • XGBoost • KNN Working with
Text • BlazingText • Supervised • Unsupervised* Recommendation • Factorization Machines Forecasting • DeepAR Topic Modeling • LDA • NTM Amazon SageMaker에서 제공하는 Built-in Algorithms Sequence Translation • Seq2Seq* Clustering • KMeans Feature Reduction • PCA • Object2Vec Anomaly Detection • Random Cut Forests • IP Insights Computer Vision • Image Classification • Object Detection • Semantic Segmentation Regression • Linear Learner • XGBoost • KNN https://docs.aws.amazon.com/sagemaker/latest/dg/algos.html

Model Deployment: Built-in Algorithms xgboost linear-learner PCA DeepAR BlazingText Image
classification … Object Detection Built-in Algorithm Images Elastic Container Registry

Model Deployment: Bring Your Own Model

def model_fn(model_dir): # Loads a model for inference model =
Your_Model() return model def input_fn(input_data, content_type): # Deserializes the input data return decoder.decode(input_data, content_type) def predict_fn(input_data, model): # Calls a model on data deseralized by input_fn return model(input_data) def output_fn(prediction, content_type): # Serializes predictions from predict_fn return encoder.encode(prediction, content_type) https://github.com/aws/sagemaker-inference-toolkit BYOM Deployment: Inference Handler Script https://github.com/tensorflow/serving def input_handler(data, context): # Pre-process request input before it is sent to TensorFlow Serving REST API if context.request_content_type == 'application/json': pass if context.request_content_type == 'text/csv': pass def output_handler(response, context): # Post-process TensorFlow Serving output before it is returned to the client response_content_type = context.accept_header prediction = response.content return prediction, response_content_type

A fully managed service that enables data scientists and developers
to quickly and easily build machine-learning based models into production smart applications. Amazon SageMaker

최초의 완전 통합 개발 환경 (IDE) 학습 모델 구축 및 협업 SageMaker Notebooks SageMaker Pipelines 완전 자동화된 머신 러닝 워크플로 구축 학습 모델 훈련 및 검증 SageMaker Training Job One-click 배포 , 모델 모니터링 및 고품질 유지 SageMaker Endpoints 학습 모델 최적화 및 다중 알고리즘 튜닝 SageMaker HPO

Let’s build a Machine Learning powered Application

Building a Visual Search Application https://aws.amazon.com/blogs/machine-learning/building-a-visual-search-application-with- amazon-sagemaker-and-amazon-es/

Visual Search Amazon Kinesis Data Streams Client AWS Lambda Amazon
S3 bucket with images Amazon API Gateway Amazon SageMaker Amazon OpenSearch Service Notebook Model Models bucket SageMaker Endpoint Train

Demo: Visual Search Application https://tinyurl.com/53yubfzw

Your Dream Comes True

문제 해결에 가장 적합한 AI 서비스 선택 문제 해결을 위한
ML 모델 개발 ML 모델 배포 및 모니터링 서비스 개발 문제 정의 AI 서비스 활용 Business Problem – ML problem framing SageMaker 활용 서비스 출시 How to choose the most relevant advertisement? How to tag millions of images in fast, cost-efficient way? How to build our own ML model? How to customize labels? Wrap-up Our Business Problem

AI/ML Stack을 활용한 비즈니스 문제 해결 과정 문제 해결에 가장
적합한 AI 서비스 선택 문제 해결을 위한 ML 모델 개발 ML 모델 배포 및 모니터링 서비스 개발 문제 정의 AI 서비스 활용 Business Problem – ML problem framing SageMaker 활용 서비스 출시

Put machine learning in the hands of every developer Our
mission at

References • Image Insights - Automatic Image Tagging & Analysis
System: https://tinyurl.com/kxydd2wn • Visual Image Search with Amazon SageMaker and Amazon ES: https://tinyurl.com/53yubfzw • Amazon Rekognition으로 나만의 컴퓨터 비전 모델 만들기: https://youtu.be/tioJru6b_4M • SageMaker Immersion Day Workshop ✯✯✯ : https://sagemaker-immersionday.workshop.aws/ • SageMaker Examples (100+) ✯✯✯ : https://sagemaker-examples.readthedocs.io/ • Advanced SageMaker Workshop: https://sagemaker-workshop.com/ • Amazon SageMaker 데모 (2020-03-25) ✯✯✯ : https://youtu.be/miIVGlq6OUk • How Startups Deploy Pretrained Models on Amazon SageMaker: https://tinyurl.com/4nwy9bd5 • Deploy trained Keras or TensorFlow models using Amazon SageMaker: https://tinyurl.com/nf2rwrw9 • Host multiple TensorFlow computer vision models using Amazon SageMaker multi-model endpoints: https://tinyurl.com/6w8f4862

Amazon SageMaker로 딥 러닝 기반 이미지 검색 서비스 만들기

Amazon SageMaker로 딥 러닝 기반 이미지 검색 서비스 만들기

More Decks by Sungmin Kim

Other Decks in Programming

Featured

Transcript