Cartoonizer

• 

  home

• 

  output

Inspiration

The project started from a simple question: Could image filters be made smarter and more creative using ML? Most cartoon filters feel flat or repetitive. I wanted to build a system that transforms regular photos into stylized cartoon images using actual deep-learning-based edge and color extraction.

What it does

Cartoonizer takes an input image, processes it through a series of filters, enhances edges, smooths colors, and outputs a cartoon-like version. The goal is to keep the structure of the image while giving it a stylized illustrated look.

How we built it

I implemented the logic using Python, OpenCV, and basic ML-inspired filtering steps. The workflow includes bilateral filtering, adaptive edge detection, and layer blending. The system runs locally and does not require heavy models, so it stays fast and lightweight.

Challenges we ran into

Tuning edge detection to avoid overly sharp or noisy output took multiple iterations.

Balancing performance and image quality was tricky.

Some images behaved inconsistently due to lighting variations, so preprocessing had to be adjusted.

Accomplishments that we're proud of

Achieved stable, repeatable cartoon-style output.

Kept the processing pipeline efficient without needing GPU-heavy models.

Built a clean, understandable codebase that others can extend.

What we learned

Small changes to filter parameters can dramatically change the visual output.

Preprocessing is critical for consistent results across a wide range of images.

A lightweight approach can still produce visually strong results.

What's next for Cartoonizer

Adding a simple UI so users can upload and preview images directly.

Experimenting with AI-based style transfer models for more artistic looks.

Offering multiple cartoon styles instead of a single output mode.

Built With

• django
• python