https://github.com/hiyali/pilgen

The aim of this repository is to generate datasets (image & its label) for OCR training.
https://github.com/hiyali/pilgen

dataset generate image-label ocr uighur uyghur

Last synced: about 1 month ago
JSON representation

The aim of this repository is to generate datasets (image & its label) for OCR training.

• Host: GitHub
• URL: https://github.com/hiyali/pilgen
• Owner: hiyali
• Created: 2020-02-21T12:36:40.000Z (almost 5 years ago)
• Default Branch: master
• Last Pushed: 2020-11-22T10:27:39.000Z (about 4 years ago)
• Last Synced: 2024-03-01T04:38:04.123Z (10 months ago)
• Topics: dataset, generate, image-label, ocr, uighur, uyghur
• Language: Python
• Homepage:
• Size: 506 KB
• Stars: 7
• Watchers: 3
• Forks: 0
• Open Issues: 1
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Pilgen - (WIP: optimizing)

Python Image-Label dataset Generator for OCR

## Generate

```bash

python3 generate.py --lang ug --count 100 --out-dir data/

```

This command will output `100` images into folder `data/images/`, filename pattern is `'word_{}.jpg'.format(line_num)`, exmaple:

```

data/images/word_1.jpg

data/images/word_2.jpg

...

data/images/word_100.jpg

```

and a `gt.txt` file, its content pattern is `'{}\t{}'.format(filepath, word)`, like below:

```

data/images/word_1.jpg	ئانا

data/images/word_2.jpg	تىلىم

...

data/images/word_100.jpg	گۈللە

```

## Supported languages

* [x] ug - Uyghur (Uighur)

* [ ] other langs may will come

## FAQ

1. How use your own corpus?

Ref: [#2](https://github.com/hiyali/pilgen/issues/2)

2. Uyghur words are separated in image?

Ref: [#2](https://github.com/hiyali/pilgen/issues/2)

## Test

```bash

python3 test.py

```

## Develop environment

* Ubuntu 18.04.1

* Python 3.6.9

## Author

Salam Hiyali

## Contribute

> Feel free

## License

MIT