Many casualties happen in warehouses. So we came up with an idea to detect if any human is in range of a machine (Eg. Forklifts.). We have a model that detects if a human is wearing safety helmet or not, and this model can be implemented on machines to check if a human exists in its range, to prevent casualties, especially of human life.

What it does

The application takes an image and processes it to check whether the person in the image is wearing a helmet or not. We can implement our algorithm in the controller box. So if a human exists in the range of a machine, and isn't wearing a helmet, we can code the machine to turn on the red indicator. Similarly if a human exists and is wearing a helmet we can make machine turn on yellow indicator. This can be done to make the controller aware of presence of a person. Similarly if the controller itself isn't wearning a helmet, we can code machine to not function.

How we built it

The model was built using pyTorch and connected to the web page using Flask as the backend. The web app was deployed on Microsoft Azure using docker containers.

Challenges we ran into

Connecting the model with the flask backend was very challenging

Accomplishments that we're proud of

We achieved an accuracy of 96%, which is something we are really proud of

What we learned

We learnt about training models using pyTorch and exporting these models to connect with the flask backend

What's next for Third Eye

Improve the accuracy and also build a better interface. Another feature would be to add automated messages incase the person is not wearing an helmet.

Share this project: