Inspiration

The need for better dataset inputs for ML models. As they say, garbage in garbage out.

What it does

Removes background from images using machine learning (see previous page)

How we built it

Tested various models to find the most accurate model. Realized that we need to crop the objects so we don't have misidentification.

Challenges we ran into

Monochromatic images gave the artificial intelligence problems in detecting subject vs background.

Accomplishments that we're proud of

We're proud that our project is capable of very accurately separating the subject of an image from its surroundings, even with images that supposedly have multiple subjects.

What we learned

How to refine our machine learning algorithm's inputs through careful testing and revising.

What's next for Back Off

Apply background removal to videos in real time.

Built With

Share this project:

Updates