{"id":1058,"date":"2020-05-13T20:48:48","date_gmt":"2020-05-13T19:48:48","guid":{"rendered":"http:\/\/www.klaushaller.net\/?page_id=1058"},"modified":"2022-03-26T10:09:26","modified_gmt":"2022-03-26T09:09:26","slug":"tutorial-first-steps-in-image-classification-using-python-jupyter-notebook-and-gluon-cv-on-aws-sagemaker","status":"publish","type":"page","link":"https:\/\/www.klaushaller.net\/?page_id=1058","title":{"rendered":"Tutorial: First Steps in Image Classification using Python, Jupyter Notebook and Gluon-CV on AWS SageMaker"},"content":{"rendered":"\n

Klaus Haller, 13.5.2020<\/em><\/p>\n\n\n\n

Click here <\/a>to download the tutorial as a pdf file – or scroll down to read it online…<\/p>\n\n\n\n

1.    Aim of this Computer Vision Introduction Tutorial<\/a><\/h1>\n\n\n\n
\"\"<\/figure><\/div>\n\n\n\n

Deep Learning, Analytics, Computer Vision \u2013 these are some of today\u2019s hot topics in the IT industry. These are key technologies for innovation. Universities and big tech companies put high efforts in research and engineering in these areas. However, even individual developers can enrich their applications with such artificial intelligence and computer vision technology if they on existing technologies, tools, and libraries.<\/p>\n\n\n\n

This short tutorial explains you one way to achieve this. We use Python and Jupyter Notebook running on Amazon SageMaker to \u201cimplement\u201d image classification based on available, pretrained neural networks within one to two hours. In other words: You will learn that you do not need any research, no Ph.D., and not two years and a big team to incorporate such features in your applications.<\/p>\n\n\n\n

The first half of the tutorial is about navigating the AWS web console, whereas the second part covers the code to get your first images classified.<\/p>\n\n\n\n

2.    Amazon SageMaker \/ Amazon Sage Maker Studio<\/a><\/h1>\n\n\n\n

SageMaker is Amazon’s solution for machine learning. It comes with one big benefit if you start learning computer vision with SageMaker: You do not have to do any installations on your local laptop. When you have your AWS login, it takes you less than five minutes and you can start with development work. And it is free, at least for the first two months and if you use it moderately, i.e., you stay away from expensive AMIs or extra options.<\/p>\n\n\n\n

To continue, you need an AWS account and you have to be signed in to your AWS account. Alternatively, you have to have a local Jupyter Notebook installation on your computer. In this case, you can continue with section 3 \u201cUnderstanding Juypter Notebook\u201d.<\/span><\/em><\/p>\n\n\n\n

xxWhen you are signed in to your AWS account and you are in the AWS services page, you type in \u201cSageMaker\u201d and select and click on the service.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Then, you click on \u201cAmazon SageMaker Studio\u201d, which is Amazon\u2019s development environment.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Next, you create a new user for SageMaker Studio. Therefore, click on \u201cAdd user\u201d.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Select \u201cQuick start, add a user name, and select \u201cCreate new role\u201d.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

A mask shows up for creating an IAM role. If you are in a test environment, just accept by clicking on \u201cCreate Role\u201d.<\/p>\n\n\n\n

Now you can create the SageMakerStudio environment by clicking on \u201cSubmit\u201d.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

A couple of times, I had issues with the submit button. I could click it, but there was no reaction. In such a case, you are most probably a person with some experience using AWS. You might want to choose the \u201cstandard setup\u201d option and choose an existing VPC or subnet.<\/span><\/em><\/p>\n\n\n\n

A little bit later (it might take a while), you see the following page and you can select \u201cOpen Studio\u201d.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Now, click the \u201cCreate a notebook\u201d button.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

You get to the following screen:<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Congratulations. You are now ready to go! The next section will give you a brief introduction in Juypter Notebook and how to use it as a development environment.<\/p>\n\n\n\n

3.    Understanding Juypter Notebook<\/a><\/h1>\n\n\n\n

Jupyter Notebook is a development environment used by many data scientists. Compared to your typical Java Eclipse IDE, there are important differences:<\/p>\n\n\n\n