Template Chinese Handwriting OCR Extraction
Publicada el 2026-07-24
Descripción de la oferta
I have a fixed, multi-field form that clients fill in by hand and I need a small program that can read the scanned JPG or PDF (Chinese handwriting) and return the content of each handwritten field. The template may changes; however, there are many separate zones—including Name, Address, Date of Birth, 行動電話, 險種名稱 and any other handwritten box on the page—that must be recognised automatically. Rather than hard-coding coordinates, I want to call the software with a list of field labels (for example “name” and “address”) and receive the recognised text for each label in a clean machine-readable format such as JSON or CSV. Internally you are free to use OpenCV, PaddleOCR, Tesseract, TensorFlow, PyTorch—or any combination—so long as the handwriting recognition works reliably on Chinese characters and can be extended if new labels are added later. Deliverables: • A runnable script or small service (Python preferred) that accepts a single image/PDF upload and returns the extracted values. • Clear instructions for installation and basic retraining or fine-tuning should accuracy need improvement. I will provide the blank template, a handful of completed samples for testing, and the definitive list of field labels. Once everything is working to at least 70 % accuracy on those samples, the project will be considered complete.
Skills
Fuente original: freelancer