Inspiration
Current applications rely on government agencies and owners of establishments like malls to implement crowd monitoring devices or deploy manpower to monitor crowds. This is costly and only a few establishments can do this. Therefore, we were inspired to create an application to monitor crowd sizes useing crowd sourced data from people who visited these public places.
What it does
It collects data from images taken from people's smartphones to detect crowd size. The data is uploaded to a centralised server and visualised on a map, allowing people to avoid crowds and stay safe during the COVID-19 pandemic.
How I built it
GUI is built with Vue and JS. Firebase is used to store the data such as images and crowd sizes. PyTorch was used to train the model and firebase was used for inference
Challenges I ran into
The AI was stuck at 33% accuracy and I couldn't fix it.
Accomplishments that I'm proud of
The application works.
What I learned
I learnt how to make a crowd counting AI.
What's next for Squeezy
Display the user’s current location on the map Path-finding algorithm to trace a path to the destination that avoids large crowds Use a larger dataset such as NWPU crowd Use a dataset that is more contextualised to COVID-19 (a lot of the photos in ShanghaiTech show giant crowds which do not happen during COVID-19)
Log in or sign up for Devpost to join the conversation.