Crossing the road entity detection

Inspiration

Current context: The current set up of the traffic lights crossing for the elderly pedestrian is by having their ID tap on the card reader above the button.

Problems targetted:

Elderly people may have forgotten to bring along their ID with them or have lost them or are unable to get near enough to the traffic light to tap due to crowd of inconveniences they have.
Only selected traffic lights currently have this function, not all. Cost issue to have this current card tapping system implemented island-wide. But with the AI model, it can be more easily implemented and scaled up in a more cost efficient manner.
Reduces the cost and manpower involved in maintaining and servicing of the current card tapping system, as well as issuing and re-issuing senior citizen cards with the card tapping function.
Includes the handicapped members of society into the equation, without additional costs, like issuing of cards for tapping purposes.

To ease the process, we came up with the idea of image recognition to detect elderly and increase the duration time based on facial recognition. We also thought of the handicapped people crossing the road who would need more time. By having image recognition, it will ease the process of the targeted audience in need to cross the road. In addition, helps with potential long-term cost savings, as well as ease of scaling the implementation of such a rollout.

What it does

The script will detect any elderly or wheelchair in the image. If either is detected, time duration will increase. Else, the timer will have a default countdown.

How we built it

We built it using Python and AWS Rekognition service

Challenges we ran into

Firstly, we had to ensure that the elderly and wheelchair detection accuracy is high and reliable. Secondly, we had to optimise the performance of the script to work efficiently with large volumes of images. Secondly, finding out and explicitly fleshing out in quantifiable terms the industry value our project brings to the table and the wider society

Accomplishments that we're proud of

We are proud of developing a system that can detect elderly and wheelchair users in images with a high degree of accuracy. We are also proud of being able to integrate our script with AWS Rekognition, which is a powerful image recognition service. Furthermore, to be able to create an AI model that not only helps with the cost-saving and monetary objectives, but also does good for the society and general population.

What we learned

While building this project, we learned a lot about the AWS Rekognition service and its capabilities. We also learned about different techniques for image recognition and object detection.

What's next for Crossing the road entity detection

More handicapped objects to be detected with training, and adding on voice command for the blind people who cross the road, as well as refining the duration added for crossing the road based on the nature and kind of people crossing