Big Bowler Brand

Inspiration

We started wanting to develop a tool that governments can use to detect who is wearing a bearded disguise. We pivoted upon realising we were missing the focus of the hackathon. We ended with a tool that can generate a bearded vehicle on a cobbled road.

What it does

Generate images of bearded vehicles.

How we built it

We used a COLAB notebook, and a huggingface stable diffusion model which has been fine-tuned on custom-made bearded vehicles. We used another huggingface stable diffusion model for upscaling at the end.

Challenges we ran into

• We severely misunderstood the brief for the first few hours.

Accomplishments that we're proud of

• We built a model that fully generates images and upscales them.

What we learned

• How to use stable diffusion.
• How to fine-tune huggingface models
• How AI-upscaling works

What's next for Big Bowler Brand

• We will go wherever the wind may blow
• Use inpainting to create a mask of the hoods of the cars.

Built With

• huggingface
• python