Document Classifier

The classifier recognizes invoices, receipts, passports, identity documents, contracts, bank statements, and resumes, and it shows a confidence score for every result. If the model is unsure, the score makes that uncertainty visible instead of hiding it.

Yes. A lightweight classification engine bundled with the page runs entirely on your device, so a document image or PDF never leaves your browser. This keeps sensitive paperwork private and lets you classify documents without an account.

Both image files and PDFs are accepted. PDFs are rendered into pixels before classification, so a scanned PDF contract behaves exactly like a photo of a contract when the model reads it.

OCR models for English, Nepali, Hindi, German, French, and Spanish are available, and the classifier works best when the document text matches one of those languages. Documents in other languages may still classify, but with lower confidence.

The score is the model's probability that a document belongs to a particular category, shown as a percentage. Higher percentages mean the match is more certain, while low scores across all categories usually indicate an unusual or mixed document type.

Accuracy depends on image quality, layout, and how closely the document matches typical templates. Clear, well-lit, high-contrast scans of standard forms classify best, while heavily styled or low-resolution documents may be labeled with a lower confidence score.