Pickturam

• 

  Figma Prototype

• 

  Albums Page

Inspiration

We were inspired by the power of multi-modal large language models. In particular, we were interested in Google's Gemma 3 models, especially with the use of quantization to reduce the size of the models.

What it does

The app utilizes a user's image gallery and leverages the power of a large language model to provide suggestions on how to improve the image.

How we built it

We used Flutter to build the app and utilized Gemini to analyze images.

Challenges we ran into

It was difficult to run a gemma model on a mobile iOS device. It was also difficult to get the app to run natively on the mobile iOS device and required extensive debugging. We had a difficult time utilizing Flutter to modify images.

Accomplishments that we're proud of

We are proud of being able to create a functioning app. It was also great to learn about using the cloud to fine-tune a large language model and utilizing APIs. One of the most exciting aspects of this hackathon was learning how to troubleshoot and resolve a wide variety of issues.

What we learned

We learned a lot about using Figma to design a template that can be implemented in code. We also learned about the use and implementation of large language models on mobile systems. We had hands-on experience in the process of debugging various issues and how to develop software and resolve problems in extreme situations.

What's next for Pickturam

Custom-tailored image modifications (user can specify which adjustments they want with a simple tap), automatic image modifications, image sorting, running a large language model locally.

Built With

• android
• android-studio
• dart
• figma
• flutter
• gemini
• ios
• vertex