Street Smart

• 

  Annotated image that uses HuggingFace AI to detect cars.

• 

  Example of our web app

Inspiration

Looking through Singapore's data.gov’s webpage for possible datasets to use and we saw traffic camera API and thought about the limitations of Google Maps and how we could provide solutions to solve these constraints

What it does

Enable commuters to obtain instant updates on congestion levels to ensure a smoother commute on the roads.

How we built it

With the help of Hugging Face’s zero-shot object detection AI model on real-time traffic images, we were able to track the number of vehicles on the roads to determine the degree of congestion.

Challenges we ran into

Overlap of car detection when using HuggingFace AI, as the AI detected multiple cars in a single car. We overcame this by increasing the score threshold for a car to be detected as a car and thus reduced the overlaps and duplicates

Accomplishments that we're proud of

Successfully integrating HuggingFace AI into our backend to enable AI car detection.

What we learned

How to work in a team and communicate our ideas to one another

What's next for Street Smart

Give Expressway recommendations based on the road conditions and have more accurate car detection with our model

Built With

• data.gov.sg
• flask
• huggingface
• python
• react
• vite