What inspired me
One day, I attended an ISEF event with my father, who was one of the main organizers. Among all the incredible student projects, one stood out and changed my perspective on computer science forever: a robotic hand controlled by human movement, built by a student named Kareem. The device mirrored each gesture through wires and sensors—it felt like magic. As a child, I was mesmerized. That experience planted a seed. From that day on, I made a promise to myself to build something even greater. Kareem, who is now a graduate of Minerva University and works at Meta, became a role model I still look up to. My love for tech deepened when I began fixing my own laggy tablet. I refused my IT-specialist father's help and spent hours diagnosing and understanding every glitch. That obsession with the “why” behind problems eventually led me to artificial intelligence.
What I learned
This project taught me far more than just coding skills. I learned how to work with OpenCV, DeepFace, and face recognition libraries from scratch—completely self-taught. I developed strong problem-solving skills by debugging tricky issues, and I gained confidence in real-world testing. Most importantly, I discovered how technology can be used to improve lives in deeply personal and emotional ways. I also learned to define success not by how accurate an algorithm is, but by how useful and meaningful it is to people.
How I built the project
I started by learning Python through Harvard’s CS50 course and continued self-studying every day. After grasping the basics, I began experimenting with face recognition models using OpenCV. I spent months coding, testing, and refining the project—often for 2–3 hours after school. I trained the system to recognize stored faces, added a text-to-speech feature to announce names, and built a working prototype of AI glasses for the visually impaired. Eventually, I combined both face detection and voice response in a single Python application, ensuring it worked in real-time under various lighting conditions.
Challenges I faced
There were many setbacks. Some photos failed to load due to memory limitations. The face recognition algorithm often failed in low light or made incorrect identifications. Getting the system to work consistently in real time on low-spec devices was a huge challenge. I also struggled with compatibility issues between Python versions and heavy libraries like torch and dlib. But every failure taught me something. I kept refining the algorithm, optimizing memory usage, and testing with real users until the system became stable and usable.
Log in or sign up for Devpost to join the conversation.