The Problem Most existing dot-to-dot products are static, generic PDF files. If you want a dot-to-dot of your dog, your favorite landscape, or a child’s drawing, you're out of luck. The current options prioritize volume over personalization, missing the demand for custom digital goods seen on platforms like Etsy.
The Solution / How it Works I developed a backend image processing pipeline that uses edge detection and simplification algorithms to extract the core contours of any uploaded image. This process drastically reduces the number of points required while preserving the recognizable form.
The algorithm then intelligently places a sequence of numbered dots along the optimized contours. This isn't just a filter; it's a structural transformation to generate a genuine, solvable dot-to-dot puzzle.
Key Features / Tech Customization: Upload any JPG/PNG and get a personalized puzzle.
Vectorization: Outputs high-quality, print-ready PDF/Image files.
Performance: Image processing is handled server-side for rapid generation.
Tech Stack: Built with Python (for image processing algorithms) and React/Next.js for the frontend interface.
I'm keen to hear from the community: What other image processing features would make this a killer utility?
Franklinjobs617•3mo ago
A common question I anticipated was the complexity. It’s tricky to balance enough detail for the final image to be recognizable, while keeping the dot count manageable for a fun puzzle. The contour simplification algorithm handles this by focusing on areas of high contrast and ignoring fine texture noise.
A technical detail: I spent significant time optimizing the dot sequencing step. It’s not simply linear. For complex shapes, the algorithm dynamically determines the shortest Eulerian path that touches all necessary contour points, ensuring the final numbered puzzle remains logical and solvable for the user.
I know the printable market is saturated, but I believe the focus on hyper-personalization is the necessary differentiator. I'm especially interested in feedback on the algorithm's performance on photos versus line art. Let me know what you think!