Posts

Over the next year, I will be completing a research project exploring how biased training data effects the machine learning algorithm of self-driving cars. This project, which I’ll detail in a later post, involves deep neural networks (DNNs), computer vision, linear algebra, and more. Is this project a massive undertaking? Yes. Will that stop me? Absolutely not. Let’s get started. Neural Network Basics First, what is artificial intelligence (AI)? AI is a field of computer science pertaining to programming computers so that they demonstrate human-like intelligence. ...

In my first Capstone post I gave a rundown of basic AI terms and how to use Tensorflow to create your own machine learning (ML) script. I’m using Tensorflow to write a script that can process images of roads and determine whether there is an obstacle in the road or not. Since I’m studying self-driving cars, I’d like to see if using biased data to train an object detection script affects the model’s performance. To do this, I’m using a training dataset of clear, bright road images to train a neural network and then testing that script with corrupted images of roads. If the model can’t recognize obstacles in images that are corrupted, then the biased training dataset did have an affect on the model. ...

Bash, or the Bourne-Again SHell, is named after the creator of the Unix shell Stephen Bourn. Bash is a command language interpreter that can execute commands and process text files as input. Shell scripts are, basically, files containing a series of commands. You can schedule Bash scripts to run at certain times to automate tasks for you. If you are using a Linux machine, you can write and execute Bash scripts in terminal. Through the terminal you can use your preferred text editor (nano, vim, etc.) to write your script and save it with the extension .sh. If you are using a MacBook, like me, you can use the terminal or Visual Studio Code to write Bash scripts- just save them with the .bash extension. ...

Kubernetes provides a framework to orchestrate multiple containers, verses working with one container at a time. To practice managing containers I opted to build a Kubernetes cluster on Raspberry Pis using Rancher. Rancher came about soon after Docker introduced containers to the market, and provides an open-source platform for enterprises to easily use Kubernetes in their environment. K3S is a Rancher product that allows you to deploy a Kubernetes cluster using a single binary file thats less than 40MB. K3S is ideal for Raspberry Pis, since traditionally building a Kubernetes cluster on Pis could be too much for them to handle. In addition to K3S this is what I used to build a simple Kubernetes cluster in my home lab: ...

Introduction to Terraform Terraform by Hashicorp is a “tool for building, changing, and versioning infrastructure safely and efficiently”. Terraform works using infrastructure as code, meaning that build a blueprint of the resources you need in a configuration file (with the .tf extension) and Terraform builds that infrastructure for you. The Terraform package itself is a single binary file, easily downloaded, and is used to build/edit/destroy your resources. In contrast to older versions of infrastructure build tools like Packer, Terraform allows you to create infrastructure across multiple cloud providers and is “smart” enough to know how many more machines to create in context of what you already have. For example, if you have 5 servers on Azure, but you want 10, Terraform will build 5 more for you to bring you to a total of 10 versus building 10 more servers and leaving you with 15. ...

Practicing installing/updating/building machines is much more fun when you aren’t running the risk of ruining an entire computer. To give myself a safe environment to work in, I have been using VMware’s vSphere in our home lab for many of my computer projects. vSphere is a suite of virtualization products that allow you to create and manage VMs. This allows me to work on VMs that can run any OS I like, and if I something goes wrong I can just delete the VM and begin again. ...