tieubaoca
/

ArchVision-OCR-LORA-vi

Model card Files Files and versions

You need to agree to share your Hugging Face information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this model content.

ArchVision-OCR-LORA-vi

LoRA adapter fine-tuned for PaddleOCR-VL-1.6 on Vietnamese OCR and Document Layout Recognition datasets:

Base model:
LoRA rank: 16 (alpha: 32)
Target modules: Text Decoder (18 layers, ~12.1M trainable parameters)
Epochs: 3.0 (9,166 steps)

Training Datasets:

(Vietnamese Handwriting)
(Table Recognition OTSL)
(Document scan)
(Title / Uppercase OCR)

Performance:

Overall CER: Reduced from 14.76% to 5.80% (over 60% relative error reduction)
Exact Match: Increased from 21.8% to 44.3%
Handwriting CER (UIT_HWDB): Reduced from 22.5% to 5.6%

Downloads last month: -

Model tree for tieubaoca/ArchVision-OCR-LORA-vi

Base model

PaddlePaddle/PaddleOCR-VL-1.6

Adapter

(4)

this model