Description
Extracting text from scans, invoices and IDs, traditional OCR struggles with complex layouts and skew. docTR is a high-performing document text recognition library powered by deep learning.
Its two-stage detection and recognition models keep layout structure, on PyTorch, usable in a few lines.
Detection:Regions.
Recognition:Accurate.
Layout:Structured output.
Easy:Few lines.
Its two-stage detection and recognition models keep layout structure, on PyTorch, usable in a few lines.
Features
Detection:Regions.
Recognition:Accurate.
Layout:Structured output.
Easy:Few lines.
