Frogstagram

View on GitHub

What if Instagram only allowed frog photos? That was the ridiculous question that sparked this project.

Frogstagram is exactly what it sounds like: a social photo-sharing platform where the only content allowed is pictures of frogs. Try to upload a selfie? Rejected. A sunset? Nope. Your cat? Not unless it’s a very frog-shaped cat. The twist is that I’m not manually reviewing every upload - a computer vision model handles that automatically.

The Idea

I wanted an excuse to deploy a machine learning model in a real production environment, but I didn’t want to build yet another generic image classifier demo. So I thought: what’s the most absurdly specific use case I could come up with? Frog detection it was.

The concept is simple but it touches a lot of interesting technical challenges: user authentication, image storage, running ML inference at scale, and building a responsive frontend that doesn’t feel like a tech demo.

How It Works

When you upload a photo, it doesn’t just go straight to the feed. First, it gets sent to an AWS Lambda function running a YOLOv11 object detection model (packaged in a Docker container via Amazon ECR). The model analyzes the image and decides: frog or not frog?

If it detects a frog, congratulations - your photo gets stored in S3 and shows up in everyone’s feed. If not, you get a rejection message. It’s a harsh world out there for non-frog content.

The Tech Stack

The frontend is built with SvelteKit and Tailwind CSS, which made it easy to create something that actually looks and feels like a modern social app. The backend runs on FastAPI deployed to AWS Lambda, so it scales automatically and I’m not paying for idle servers.

For authentication, I used AWS Cognito so users can create accounts and have their own profiles. All the images live in S3, which handles the storage and delivery.

Live Demo

Note: The legacy version of this project, which used vanilla JavaScript for the frontend and FastAPI with SQLAlchemy for the backend, and TensorFlow for inference, can be found at Frogstagram Legacy. The current production version is at Frogstagram Live.