Shark on the Road

Screenshot of the program in action.
Challenges we ran into. Behold! A bird.

Inspiration

Driving can be very dangerous. It can be even more dangerous if there were sharks on the road. Luckily, we have prepared a solution for this.

What it does

This program captures a video, and uses machine learning to detect the location of sharks on the road.

How we built it

This is built using two machine learning models. One for identifying potential road sharks, and one for identifying if it is indeed a shark.

Challenges we ran into

We discovered halfway through the project that unfortunately the COCO 2017 dataset which is commonly used to detect objects, does not contain images of sharks, causing them to be identified as a variety of animals and objects. To solve this, we combined this with the utilization a different classifier that was previously trained with a different dataset which includes a variety of sharks, which was used to verify that potential images of sharks were actually sharks.

Accomplishments that we're proud of

We were able to create an object-detection machine learning project with limited knowledge of the subject and the program works with somewhat accurate results.