Building a Satellite Airplane Detection System on AWS
A deep dive into deploying a computer vision research pipeline (IPGBBS + P-SIFT) on AWS EC2 using the RarePlanes dataset. Learn how to bridge the gap between research papers and real-world, deployable cloud solutions.
Introduction
As a Computer Science student, one of my most challenging and rewarding projects has been bringing a published computer vision research paper to life. I built a Satellite Airplane Detection & Tracking System, implementing the IPGBBS and P-SIFT algorithms, and integrated it with the RarePlanes dataset. But the real learning came when I moved it from my local machine to a live deployment on AWS EC2.
Why This Project?
The goal was twofold:
1. Deepen my understanding of computer vision beyond just using pre-built libraries.
2. Gain practical DevOps skills by deploying a complex, resource-intensive application to the cloud.
The Challenge: From Research to Reality
Research papers often leave out the "gritty" implementation details. My first hurdle was translating the mathematical descriptions of IPGBBS and P-SIFT into efficient, working Python code that could handle satellite imagery. Testing and validating the results against the RarePlanes dataset was another iterative process.
Deployment on AWS EC2
This was where I truly had to think like a cloud engineer. I had to:
· Select the Right Instance: Choose an EC2 instance with enough compute and memory to process the images and run the detection algorithms reasonably fast.
· Configure the Environment: Set up the entire Python environment, install all dependencies, and ensure the system was secure.
· Manage Ingress/Egress: Configure security groups to allow access to the application's web interface or API while keeping the backend secure.
Setting Up the Environment
First, I needed to install all the necessary tools and packages on my EC2 instance. This included Git to clone my repository, pip for Python package management, and other dependencies.

Caption: Setting up the environment on AWS EC2 – installing Git, pip, and required packages (total download size: 9.8 MB).
As shown above, the installation included:
· Git and Git core packages
· Python pip
· Perl dependencies
· libxcrypt-compat
Running the Application
After the environment was set up, I navigated to my project directory and launched the Streamlit application.

Caption: My Satellite Airplane Detection System running on AWS EC2. Successfully accessible at 54.161.3.105:8501 – proof that the deployment succeeded!
A Quick Lesson Learned: Initially, I ran streamlit run app.py from the wrong directory and got a "File does not exist" error. I quickly realized I needed to cd into the project folder first – a small but important step when deploying!
Once in the correct directory, the app launched successfully with:
The External URL meant my application was now live and accessible from anywhere in the world!
Key Takeaways
· The Power of Open Source: Being able to build upon existing research and datasets accelerates learning and innovation immensely.
· Cloud is an Enabler: AWS provides the tools to take almost any project from a local script to a globally accessible application.
· The "Always Building" Mindset: This project is a testament to my philosophy of learning by doing. It combined my interests in AI/ML, software development, and cloud computing into a single, tangible outcome.
· Troubleshooting is Part of the Process: The "File does not exist" error taught me to always verify my current working directory – a lesson I won't forget!
Next Steps
I'm currently deepening my AWS and Machine Learning skills through AWS Educate. My next goal is to refine this system, perhaps adding a simple frontend and optimizing the detection pipeline for better performance and cost-efficiency on the cloud.
Conclusion
My journey from an idea to a live deployment on AWS has been the most effective classroom. I encourage every developer—especially students—to build, deploy, and share their projects. The learning you gain from taking something from your IDE to the internet is invaluable.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article