When we began this project at IDEAHacks, we wanted to build something that would be challenging for us while still having a clear idea of what had to be done to build it. We decided upon a robot that could detect whether an area was messy when we saw all the trash and wires around us.

What it does

This robot detects whether an area is messy or clean based on an initial calibration scan for what qualifies as clean.

How we built it

We used Google's computer vision API once we realized that the Raspberry Pi could not support a model that we trained ourselves. On the hardware side of things, we used a Raspberry Pi along with a Raspberry Pi camera as our central components and used several servos in order to allow the robot to take a wide scan of its area.

Challenges we ran into

We fell into several bottlenecks along the way. At first, we were unable to set up the Pi due to a lack of a dedicated monitor or ethernet cable to let us figure out its IP address. After that, we trained a model for detecting whether a room is messy or clean using Google collab, but when we moved our model to the Pi and ran it, we realized that the Raspberry Pi had an insufficient amount of processing power to run our model.

Accomplishments that we're proud of

We started a project two days ago. And now, two days later, we have a project to show for it. It may not be exactly as ambitious as we initially aimed, but we are proud that something came out of our work.

What we learned

One of our major takeaways from this project was a basic understanding of computer vision and Google cloud APIs. One of our members had no experience with hardware coming in, and now he does. And most of all, we learned the value of friendship in our small team of two members.

What's next for WeClean

We won't be continuing with this particular project because we don't have ownership of its parts. But, what we've learned from this experience will help us in our future projects, especially ones that involve artificial intelligence and hardware.

Built With

Share this project: