Last year, 240000 illegal pilgrims where busted, thanks to the efforts of national army guards. The much appreciated human efforts comes with unfortunate errors and missed potential criminals. This makes us realize there is a big chunk of the actual illegal immigrants that remain untapped due to human limitation.
What it does
Our solution will be used for real-time detection of illegal pilgrims. Having 3 Artificial Intelligence professionals in the team, our AI model consists of the three main components:
- detect faces using CNN
- recognize faces using one-shot learning
- map associated data from Ministry of Interior Database
How I built it
The main part of the project was done utilizing the technique of OneShot learning that have been developed recently by MIT AI research lab. This is the only model that requires a single image for each person to recognize them with high accuracy compared to traditional facial recognition techniques. The project can also be tuned to detect opposite crowds and lost pilgrims.
Challenges I ran into
Limited processing power and GPU power of our laptops as well as lack of data
Accomplishments that I'm proud of
Being the only team that have combined both techniques to solve this tedious and time-consuming issue with very low latency.
What I learned
What's next for C-052-Dragonfly
The project can also be tuned to detect opposite crowds and lost pilgrims. It can also be used for headcount with very high accuracy in comparison to other models.
Log in or sign up for Devpost to join the conversation.