Union Transformer Vietnamese OCR (UniVietOCR)

Introduction

UniVietOCR is an end2end vietnamese text recognition with pre-trained models transformers.

UniVietOCR

Wrap Encoder-Vit4x4 (from Union14M) and Decoder (from VietOCR)

UniVietOCR is a Vietnamese text recognition model composed of Vit finetuned on Union14M-L (Vit4x4) and Transformer decoder pretrained on Vietnamese dataset (VietOCR)

Code: coming soon 😸

Performance

Outperforming VietOCR

Reference

1. https://github.com/Mountchicken/Union14M
2. https://github.com/pbcquoc/vietocr
3. https://github.com/microsoft/unilm/tree/master/trocr