Inspiration
This project was inspired by one of our partners who manages dozens of recycling centers. In a center, each container contains waste of the same nature: iron, plastic, green waste, etc. The users put waste in the containers as long as they are not full. The replacement of full containers is managed on demand.
This hackathon was the opportunity to propose a solution for the automatic monitoring of the filling of the containers. Thus, a central service can organize the replacement of containers according to their filling speed. It is thus possible to optimize the route of the trucks in charge of emptying the containers. In the long term, this optimization should make it possible to limit the number of trucks and the number of kilometers traveled.
In addition, by anticipating the filling of containers, it is be possible to offer empty containers to users more quickly. This will avoid asking users to come back at another time to drop off their waste.
The data of the filling of the containers through time will also make it possible to foresee the filling of containers.
What it does
Our solution is based on the use of Meraki cameras. We were lucky enough to be able to place a camera above a container to carry out our project. The images are retrieved and analyzed to determine the fill rate.
For this project we display the fill rate and for a future use (prediction) we used the person counting possibilities offered by the camera. The number of users per hour could for example be used to take into account a priority in the replacement of containers.
The architecture of the solution is described in the attached presentation.
How we built it
We placed a camera above a container in a recycling park. Thanks to the Meraki API we get images and number or people every 5 minutes. We built a AI model to get the filling rate. The result of all these python scripts is image files and csv files. We developped a small tool to visualize the images of containers and simultaneously the filling rate and the number of people around the container.
Challenges we ran into
First challenge was to build a AI model to get the filling rate of a container. Next, we are a team of three people. it was easy to divide the work but next time we will try to be more numerous.
Accomplishments that we're proud of
The model to estimate de filling rate even if we are sure that a embedded solution on the camera could be possible.
What we learned
Teamwork and friendship under pressure
What's next for Automated Recycling Containers Monitoring
More deep learning to make prediction thanks to Meraki camera.
Log in or sign up for Devpost to join the conversation.