Inspiration
We decided to make a model that generates images of IIITk after successfully completing our first semester online
What it does
Trained on over 2000 images of IIITk campus, the StyleGAN2 model can successfully generate beautiful images of the campus
How we built it
We started of by trimming the video to extract the parts that had the campus. Using OpenCV we extracted and saved the frames. After scraping some more images from google photos we had over 2000 images of the elegant campus. We then took a model trained on images of church buildings and trained it with the campus photos for ~24 hours to get a model that could successfully generate fake images.
Challenges we ran into
Training the model on free cloud providers and not having enough computation which ended up taking most of our time.
Accomplishments that we're proud of
A model that can successfully generate images specific to iiitk campus
What we learned
How the StyleGAN2 model works and how to extract frames from a video using openCV
What's next for This IIITk campus dosent exist
this-kerala-dosent-exist? this-india-dosent-exist? Leveraging Deep Learning to make models that can accurately represent our world
Built With
- css3
- html5
- javascript
- opencv
- python
- tensorflow
Log in or sign up for Devpost to join the conversation.