Invoice Extraction for Vietnamese: some experiments

Please carefully read this README.

Train and evaluate models for Invoice Extraction.

⚠️ Before using the code, make sure you have these files downloaded:

Download URL	Save to path
Google Drive	weights/craft_mlt_25k.pth
Google Drive	weights/craft_refiner_CTW1500.pth
MC-OCR Competition 2021	data/mc_ocr_warmup_500images.zip
MC-OCR Competition 2021	data/mcocr2021_public_train_test_data.zip

Scripts:

• Prepare the 2021 MC-OCR corpus: prepare_mcocr2021.py
• (Text Recognition) PARSeq model: parseq.py
• (Text Recognition) VietOCR model: vietocr_api.py
• Scene text inference (PARSeq + VietOCR): scene_text.py

Experiments:

• [01] CRAFT + VietOCR/PARSeq inference
• [02] PARSeq inference
• [03] VietOCR inference
• [04] MC-OCR corpus example usage

I use Python 3.10.11. Any other version of Python has not yet been tested.