Build an AWS Machine Learning Pipeline for Object Detection

Name: Build an AWS Machine Learning Pipeline for Object Detection
Rating: 4.7 (103 reviews)

Use AWS Step Functions + Sagemaker to Build a Scalable Production Ready Machine Learning Pipeline for Plastic Detection

Created byPatrik Szepesi

Last updated 12/2024

English

What you'll learn

Learn how you can use Google's Open Images Dataset V7 to use any custom dataset you want
Create Sagemaker Domains
Upload and Stream data into you Sagemaker Environment
Learn how to set up secure IAM roles on AWS
Build a Production Ready Object detection Algorithm
Use Pandas, Numpy for Feature and Data Engineering
Understanding Object detection annotations
Visualising Images and Bounding Boxes with Matplotlib
Learn how Sagemaker's Elastic File System(EFS) works
Use AWS' built in Object detection detection algorithm with Transfer Learning
How to set up Transfer Learning with both VGG-16 and ResNet-50 in AWS
Learn how to save images to RecordIO format
Learn what RecordIO format is
Learn what .lst files are and why we need them with Object Detection in AWS
Learn how to do Data Augmentation for Object detection
Gain insights into how we can manipulate our input data with data augmentation
Learn AWS Pricing for SageMaker, Step Functions, Batch Transformation Jobs, Sagemaker EFS, and many more
Learn how to choose the ideal compute(Memory, vCPUs, GPUS and kernels) for your Sagemaker tasks
Learn how to install dependencies to a Sagemaker Notebook
Setup Hyperparameter Tuning Jobs in AWS
Set up Training Jobs in AWS
Learn how to Evaluate Object detection models with mAP(mean average precision) score
Set up Hyperparameter tuning jobs with Bayesian Search
Learn how you can configure Batch Size, Epochs, optimisers(Adam, RMSProp), Momentum, Early stopping, Weight decay, overfitting prevention and many more in AWS
Monitor a Training Job in Real time with Metrics
Use Cloudwatch to look at various logs
How to Test your model in a Sagemaker notebook
Learn what Batch Transformation is
Set up Batch Transformation Jobs
How to use Lambda functions
Saving outputs to S3 bucket
Prepare Training and Test Datasets
Data Engineering
How to build Complex Production Ready Machine Learning Pipelines with AWS Step Functions
Use any custom dataset to build an Object detection model
Use AWS Cloudformation with AWS Step Functions to set up a Pipeline
Learn how to use Prebuilt Pipelines to Configure to your own needs
Learn how you can Create any Custom Pipelines with Step Functions(with GUI as well)
Learn how to Integrate Lambda Functions with AWS Step Functions
Learn how to Create and Handle Asynchronous Machine Learning Pipelines
How to use Lambda to read and write from S3
AWS best practices
Using AWS EventBridge to setup CRON jobs to tell you Pipeline when to Run
Learn how to Create End-to-End Machine Learning Pipelines
Learn how to Use Sagemaker Notebooks in Production and Schedule Jobs with them
Learn Machine Learning Pipeline Design
Create a MERN stack web app to interact with our Machine Learning Pipeline
How to set up a production ready Mongodb database for our Web App
Learn how to use React, Nextjs, Mongodb, ExpressJs to build a web application
Create and Interact with JSON files
Put Convolutional Neural Networks into Production
Deep Learning Techniques
How to clean up an AWS account after you are done
Train Machine Learning models on AWS
How to use AWS' GPUs to speed up Machine Learning Training jobs
Learn what AWS Elastic Container Registry(ECS) is and how you can download Machine Learning Algorithms from it
AWS Security Best practices

Course content

13 sections • 139 lectures • 16h 30m total length

Let's look at our End Project12:28
Learn to build an end-to-end aws step functions driven machine learning pipeline for object detection, including SageMaker model development, hyperparameter tuning, batch transformation, and asynchronous polling.

Source Code for the Course0:01
Setting up IAM User8:25
Create an IAM user in AWS, avoid root access, balance programmatic and console access, apply least privilege, and securely manage administrator permissions, access keys, and login credentials.
Clarification about AWS S31:13
Getting Data for our Project4:13
Explore pre-labeled data from open images v7 for object detection, including plastic bag labels, and download via Colab using voxel 51 before uploading to AWS SageMaker.
Getting dataset Part 17:59
Getting dataset Part 24:33
Learn how to prepare and manage object detection datasets for an AWS machine learning pipeline, including organizing train and validation folders, examining labels, and creating zipped archives for efficient upload.
Getting dataset Part 32:08
discover how to locate the generated dataset in google colab, identify the train and validation image zips under content/route/51, and download the files for object detection.
Getting dataset Part 40:33
Download and verify train images.zip and validation images.zip, then upload all dataset files to AWS to begin the object detection workflow.

Create SageMaker Domain2:21
Learn to create a SageMaker domain in Studio, set up an execution role, and prepare data by uploading to a SageMaker notebook before moving to S3 and model training.
Create SageMaker Studio Notebook1:35
Launch and configure a SageMaker Studio notebook for an object detection workflow, including navigating domains and initiating first-time Studio setup. Learn about Studio's collaborative space and real-time notebook editing.
Sagemaker Notebook Updates11:58
Learning how to Stop and Start SageMaker Notebooks5:25
Learn how to create and configure a SageMaker notebook, choose a Docker image and kernel, and stop or restart notebooks to manage running costs.
Restarting our SageMaker Studio Notebook Kernel1:27
Restart the SageMaker Studio notebook kernel, rename the notebook to object detection, and save your work before uploading code and data in the next steps.
Upload and Extract Data in SageMaker7:35
Upload Open Images dataset v7 into SageMaker Studio's EFS via drag-and-drop or upload, then unzip train images and validation images into unzipped folders using Python.
Deleting Unused Files1:45
Delete unused train and validation zip files after extracting images to save storage costs and shut down running SageMaker instances to avoid charges before exploratory data analysis.

Loading and Understanding our Data8:31
Counting total Images and getting Image ids16:37
Count the number of training and validation images and extract their IDs using glob-based path patterns. Verify that the IDs correspond to files in each folder for a csv-free workflow.
Getting Classname Identifier7:25
Parse the classes.csv to map class identifiers to names, load train and validation metadata, and filter for the plastic bag object to retrieve corresponding identifiers for object detection.
Looking at Random Samples from our Dataframe11:59
Understanding Annotations6:50
Explore how annotations structure image data for object detection by interpreting image IDs, label names, and normalized bounding boxes, including multiple boxes per image and visualization plans.
Visualize Random Images Part 19:50
Visualise Random Images Part 22:16
Matplotlib difference between plt.show() and plt.imshow()0:37
Visualising Multiples Images at Once15:21
Visualize many images at once to inspect datasets using a matplotlib grid of subplots, adjustable rows and columns, and train or validation image paths for efficient data exploration.
Correcting our Function0:23
Visualising Bounding Boxes Part 14:50
Use csv data to draw bounding boxes around images for object detection, specifying image id and the plastic bag label, then visualize and prepare to automate bounding box drawing.
Visualising Bounding Boxes Part 2 (Theory Lesson)16:26
Visualising Random Images with Bounding Boxes Part 115:39
Wrong Print Statement0:10
Visualising Random Images with Bounding Boxes Part 211:06
Visualize multiple bounding boxes on random images by looping through bounding box data, converting normalized coordinates to pixel values using image width and height, and drawing red rectangles.
Read this Lesson if you have issues with Data Visualization0:30

Clean our Train and Validation Dataframes9:02
Clean and filter the train and validation frames to keep only the plastic bag label, yielding 517 train images, 986 bounding boxes, and 9 validation images with 9 bounding boxes.
Split Dataframe into Test and Train5:16
Split the df_train_cleansed into 788 train and 198 test rows (80% training, 20% testing) with train_test_split and random_state 42; note the validation set isn't used for training.
Get Images IDs11:21
Splitting IDs Theory Lesson7:46
Explore how to split image IDs for object detection by moving or copying images between train and test folders based on bounding box annotations and duplicates across dataframes.
Explanation Regarding Next video0:09
Moving Images to Appropriate Folders9:35
Move or copy images from train to test folders based on IDs, using glob and path operations. Copy when an ID appears in both train and test, otherwise move.
Count how many Train and Test Images we have4:58
Count train and test images with glob paths to verify correct placement, revealing 162 test images and 449 train images. Identify 94 duplicates where bounding boxes appear in both folders.
Verifying that our Images have been moved Properly Part 112:25
Verifying that our Images have been moved Properly Part 24:52

Using Mxnet5:44
Convert train and test data frames into LST files and then MXNet record IO for an AWS SageMaker object detection workflow, using ResNet for CNN performance.
Additional Info regarding RecordIO format0:14
Using Mxnet RecordIO14:10
Learn how to convert a data frame to lst and recordio formats for mxnet object detection, detailing image index, class ids, bounding boxes, headers, and image paths.
Correction Regarding Label width1:18
Preparing Dataframes to RecordIO format Part 112:41
Rename the label name to class name, copy dataframes to avoid warnings, add header and label width, and construct image path columns to prepare train and test data for RecordIO format.
Preparing Dataframes to RecordIO format Part 26:50
Moving Images To Correct Directory3:22
Move images to the correct image path using shuttle.move, automatically creating missing directories, and verify train and test image folders contain the expected images.
Explanation Regarding the Previous Video0:24
Verifying that all Images have been Moved Properly3:34
Read Before Proceeding to the next Lecture0:21
Creating Production .lst files (Optional)21:52
Learn how to create production-ready LST files from object detection data by consolidating multiple bounding boxes per image into one line, following MXNet conventions for train and test.

Data Augmentation Theory11:15
Augmenting a Random Image7:37
Augment a random image by transposing and flipping left-right, recalculate the x coordinates for the bounding box, and visualize the transposed image with updated bounds.
Moving Images to new Folder structure5:55
Logically reorganize multi class object detection data by moving images into a class-based folder structure, update paths and frames, and prepare to visualize and recalculate bounding boxes after transposing images.
Visualising Random Augmented Images Part 113:49
Visualising Random Augmented Images Part 211:28
Read this Lesson if you have issues visualising your images0:31
Creating Data Augmentation Function Part 117:25
Define a data augmentation function that flips images along the y axis, saves augmented images, and updates bounding box coordinates in the train or test data frames for object detection.
Creating Data Augmentation Function Part 218:54
Checking Image Counts Before running the Function2:10
In an AWS machine learning pipeline for object detection, check train and test image counts before running the function to verify data augmentation and final data frame lengths.
Correctional Video regarding our Function3:56
Augmenting Test Dataset and Creating test .lst Files6:20
Run the augment data function on the test set to create a test LST file. Verify augmented bounding boxes and flipped image paths for the test data.
Augmenting Train Dataset and Creating .lst File Part 15:10
Part 1 of augmenting the train dataset uses augment_data with the train parameter to create train.lst, verifying bounding boxes are 0 and 1 and including original and flipped images.
Augmenting Train Dataset and Creating .lst File Part 20:56
Verifying that Data Augmentation has Worked6:53

Increasing Service Quotas4:51
Installing dependencies and Packages6:31
Creating our RecordIO Files10:57
Use mxnet im2rec.py from the Apache mxnet GitHub to convert test.lst and train.lst to record IO files (.rec).
Uploading our RecordIO data to our S3 bucket11:17
Downloading Object Detection Algorithm from AWS ECR5:15
Configure the output location in S3 and download the AWS object detection algorithm from ECR, then set region, framework, and version to prepare for model training in SageMaker.
Setting up our Estimator Object7:37
Setting up Hyperparameters21:07
Set up hyperparameters for an object detection model using transfer learning with pretrained Resnet 50, defining classes, epochs, learning rate schedule, momentum, weight decay, NMS-threshold.
Additional Information for Hyperparameter Tuning in AWS0:05
Setting up Hyperparameter Ranges10:07
Set up hyperparameter tuning in SageMaker by using Bayesian search to train multiple models, exploring learning rate, mini batch size, and optimizer options like SGD and Adam for object detection.
Setting up Hyperparameter Tuner9:15
Configure hyperparameter tuning for an AWS SageMaker object detection training by setting max jobs and max parallel jobs, balancing model quality and cost with Bayesian search.
Additional Information about mAP( mean average precision)0:06
Starting the Training Job Part 18:01
Learn to configure SageMaker with train and validation data paths in S3, set data channels, choose fully replicated distribution, and launch a hyperparameter tuning job using tuner.fit with epoch adjustments.
Starting the Training Job Part 218:40
More on mAP Scores0:06
Monitoring the Training Job4:18
Monitor multiple training jobs for object detection, compare validation map scores, and tune SGD, batch size 16, and learning rate 0.0179 to identify the best model.
Looking at our Finished Hyperparameter Tuning Job8:33

Deploying our Model in a Notebook7:09
Creating Visualization Function for Inferences18:35
Learn to deploy an AWS machine learning pipeline for object detection and visualize inferences by building a visualization function that draws bounding boxes on images using SageMaker endpoints.
Testing our Endpoint Part 17:00
Testing out Endpoint Part 24:46
Test and verify the deployed endpoint by running predictions on validation images, correcting data issues, and highlighting how transfer learning enables accurate plastic bag detection with limited data.
Testing our Endpoint from Random Images from the Internet6:59
Test the object-detection model on internet images using a ResNet-50 backbone trained on 800 images (400 real, 400 augmented) and tune the threshold to 25–30%; manage endpoints to control costs.

Setting up Batch Transformation Job locally first16:30
Build a batch transformation job locally in SageMaker notebook, then implement an AWS Step Functions pipeline to produce daily object detection inferences, including bounding boxes and classes, saved to S3.
Starting our Batch Transformation Job7:18
Kick off the batch transform job by preparing sample data in the S3 bucket under images/year/month/day, and test with 4–5 images from free stock sites.
Analysing our Batch Transformation Job7:48
Visualising Batch Transformation Results15:56
Visualize batch transformation results by extracting bounding boxes, confidence scores, and class IDs from .out JSON files and drawing boxes on corresponding images in S3 and SageMaker batch tests.
Look at this lesson if you have trouble with the Visualisations1:04

Requirements

Laptop with Internet Access
AWS account
Knowledge of Python and basic Machine Learning
Spend 20-50 dollars on AWS if you want to follow along with me. Note that you can still follow along without having to pay any money

Description

Welcome to the ultimate course on creating a scalable, secure, complex machine learning pipeline with Sagemaker, Step Functions, and Lambda functions. In this course, we will cover all the necessary steps to create a robust and reliable machine learning pipeline, from data preprocessing to hyperparameter tuning for object detection.

We will start by introducing you to the basics of AWS Sagemaker, a fully-managed service that provides developers and data scientists with the ability to build, train, and deploy machine learning models quickly and easily. You will learn how to use Sagemaker to preprocess and prepare your data for machine learning, as well as how to build and train your own machine learning models using Sagemaker's built-in algorithms.

Next, we will dive into AWS Step Functions, which allow you to coordinate and manage the different steps of your machine learning pipeline. You will learn how to create a scalable, secure, and robust machine learning pipeline using Step Functions, and how to use Lambda functions to trigger your pipeline's different steps.

In addition, we will cover deep learning related topics, including how to use neural networks for object detection, and how to use hyperparameter tuning to optimize your machine learning models for different use cases.

Finally, we will walk you through the creation of a web application that will interact with your machine learning pipeline. You will learn how to use React, Next.js, Express, and MongoDB to build a web app that will allow users to submit data to your pipeline, view the results, and track the progress of their jobs.

By the end of this course, you will have a deep understanding of how to create a scalable, secure, complex machine learning pipeline using Sagemaker, Step Functions, and Lambda functions. You will also have the skills to build a web app that can interact with your pipeline, opening up new possibilities for how you can use your machine learning models to solve real-world problems.

Who this course is for:

For developers who want to take their machine learning skills to the next lever by being able to not only build machine learning models, but also incorporate them in a complex, secure production ready machine learning pipeline

Build an AWS Machine Learning Pipeline for Object Detection

What you'll learn

Explore related topics

Course content

What we are Building1 lecture • 12min

Getting Started with AWS and Getting our Dataset8 lectures • 29min

Setting up AWS SageMaker7 lectures • 32min

Exploratory Data Analysis16 lectures • 2hr 9min

Cleaning and Splitting our Data9 lectures • 1hr 5min

Date Engineering11 lectures • 1hr 11min

Data Augmentation14 lectures • 1hr 52min

Setting up and Creating our Training Job16 lectures • 2hr 7min

Analysing Training Job Results5 lectures • 44min

Setting up Batch Transformation5 lectures • 49min

Requirements

Description

Who this course is for: