Skip to content

FahNos/long-text-recognition-svtr2.1

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

40 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Text regconition for long text

This repository was build on OpenOCR frame work

I had the change in encoder part and use VNese dataset to training

This model can use for English and Vietnamese language

Results

I was implement test follow dataset of OpenOCR, and result as below:

Model LTB IC13
857
SVT IIIT5k
3000
IC15
1811
SVTP CUTE80 Avg Config&Model&Log
SVTRv2-T 47.83 98.6 96.6 98.0 88.4 90.5 96.5 94.78 Google drive
SVTRv2-S 47.57 99.0 98.3 98.5 89.5 92.9 98.6 96.13 Google drive
SVTRv2-B 50.23 99.2 98.0 98.7 91.1 93.5 99.0 96.57 Google drive
SVTRv2-1 57.67 97.5 96.4 98.5 88.8 90.9 95.8 94.7 log: Hugging face, pretrain model
  • The results show that model SVTR2-1 have best results for long text prediction, even better than SMTR model, which have reach 51.0 for long text prediction

Dataset

  • I use the dataset with the combinition between Union14M-L and VNese dataset, and the dataset distribution of ratio between width / heigh as table below:
Ratio 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Qty 180287 1102443 616312 388852 281372 183585 155346 55600 41072 35863 16112 30108 28328 26244 22112 21253 18782 17818 14994 15683 11812 10361 3704 7885 53701
Percentage 27.74% 26.02% 15.70% 9.37% 6.92% 2.93% 2.07% 1.34% 1.07% 1.01% 0.92% 0.84% 0.72% 0.67% 0.58% 0.56% 0.48% 0.44% 0.38% 0.35% 0.30% 0.26% 0.12% 0.19% 1.32%

Usage

  • You can run this model follow this demo

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published