Inspiration
We thought it would be cool to do something with computer vision, we had read a paper about the merits of a specific image to image converter and wanted to apply it to a real world problem.
What it does
Uses a computer vision machine learning model to make predictions about a photo given a training set of input photos and output photos
How we built it
We built this from scratch (i.e. didn't employ preexisting APIs) using google Colab which gave us access to a GPU in order to run GPU-intensive computations.
Challenges we ran into
Creating our own dataset, employing neural network libraries, version control, GPU access. The biggest challenge is that we did not have enough time to fully train our data set as training even a weak model would have taken over 20 hours of run time given the current specs that we had access to.
Accomplishments that we're proud of
We created a working algorithm that provides evidence of what we're trying to accomplish
What we learned
The underlying technology of neural networks and cycleGAN
What's next for JPEEG
This type of technology could be used for facial recognition of people wearing masks
Built With
- colab
- keras
- python
- tensorflow
Log in or sign up for Devpost to join the conversation.