Helmet Detection AI

Inspiration

With the rising concerns surrounding two-wheeler accidents and the importance of helmet usage, our project utilizes the power of Amazon Rekognition API to create an intelligent system that identifies and promotes helmet compliance among riders.

What it does

It detects objects from an image and checks if there is a helmet in it.

How we built it

We used the rekongnition API from AWS which is a pre-trained AI model for object detection. We then checked the example code of the API, ran the code to test it, understood the code and then remade from scratch to fullfill our use-case. We took some code from the API example code and added functions such as check_if_helmet().

Challenges we ran into

1) Optimising the processing time from 15 seconds which was challenging. 2) Integrating a RaspberryPI into the project to use it as a webcam and an indicator on wether a helmet was detected or not. Unfortunately we could not add it to our project because of time restrictions 3) Filteration of the output given by the AI was a diffult job, overall we learnt that even though the term AI sounds fancy, it is actually very difficult to do AI.

Accomplishments that we're proud of

We are proud that we could learn and make a working AI algorithm in a short time. We were also able to decrease the image processing time all the way from ~15 seconds down to ~2 seconds using methods such as numpy arrays etc.