SmolDocling Visualizer
Hey everyone! 👋
I've been playing around with SmolDocling-256M-preview and built a visualization tool that I thought might be useful for the community.
It's basically a web interface that lets you see exactly what SmolDocling is detecting in your PDFs - it draws colored boxes around different document elements (headers, text, images, tables) and can extract the images too. Super helpful for understanding how the model "sees" documents.
Here's the repo: SmolDocling-visualizer
I've found it really handy for:
- Quick demos to show people what SmolDocling can do
- Debugging when the model misses something
- Creating visual examples for presentations
It runs in Docker so should be easy to set up. Just throw in a PDF and watch the magic happen!
Feel free to use it for your own demos or presentations. I think tools like this can really help people understand the power of SmolDocling - sometimes seeing is believing..
