A Student's Journey in Applied Data Science & AI

Dániel Elek built a computer vision application in block C designed to identify and differentiate between soldiers and civilians in war zones, for drone deployment with a tablet interface.

Trained on roughly 500 images gathered from Google and other sources, the system reached 85% accuracy. Finding suitable images proved the hard part, with size and aspect ratio both needing care.

Market analysis was a substantial component of the work: competitor research, examination of existing approaches, and customer segmentation across countries.

Interpretability

Dániel put particular weight on explainable AI, integrating libraries that let a user see how the neural network reached its classification - which, for an application of this kind, is not an optional extra.

Role

He worked as market analyst, AI engineer and computer vision specialist at once, handling the technical side alongside the interface design. His own assessment is that more explainable AI methods would have improved it. He would like to develop the application further towards pinpointing the exact location of identified individuals within an image.