UnrealHawker

Inspiration

Finding seats at a hawker is always a challenge, especially during peak periods. In all likelihood, you have to physically go down and visually check if there are any seats at all. This would in turn waste time on finding space (especially for people with limited lunchtime). What if there was a convenient way to check if there are seats at a hawker and where it might be?

What it does

Using computer vision, we can identify if a table is taken based on imagery. This system allows patrons to view the occupancy of tables in hawker centres in real time and locate vacant seats via the dashboard located on the hawker centre television screen.

Flow of Events

Unreal Engine 5: Training Models Using Photorealistic Simulation for Fast Implementation

How we built it

Computer Vision Model

We created a dynamic model that tracks both plates, clutter, cups & tables, and connects the information to determine the vacancy of a table. This was implemented through openCV. If the hawker owner chooses to move tables around, our model will still be able to track correctly and accurately provide the required data

Unreal Engine 5 (UE5) Simulations

By creating a fully configurable environment in UE5, we can train our models without the need to feed it real-world images. Furthermore, we can dynamically move user controlled actors to interfere with the environment, simulating real-life circumstances

Position

Progressive Web Application and IoT

Additional sensors (pressure sensors, ultrasonic) could be used to assist in more accurate vacancy data. Implementation and iteration is simple through Progressive Web Applications. We can directly update the results of our Model to users online. This enables users to make informed decisions based on our findings. Such models can also be configured and tested using UE5.