https://github.com/aminAghayani/handwriteToText

persian/farsi handwrite to text with keras and data extraction
https://github.com/aminAghayani/handwriteToText

Last synced: 3 months ago
JSON representation

persian/farsi handwrite to text with keras and data extraction

• Host: GitHub
• URL: https://github.com/aminAghayani/handwriteToText
• Owner: aminAghayani
• Created: 2021-09-27T10:46:06.000Z (about 3 years ago)
• Default Branch: main
• Last Pushed: 2021-09-29T13:50:15.000Z (about 3 years ago)
• Last Synced: 2024-06-28T11:34:20.294Z (5 months ago)
• Language: Python
• Size: 34.2 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# handwriteToText

persian/farsi handwrite to text with keras and data extraction

This project is trying to convert persian handwriting to text; For this purpose a dataset of 32 persian characters and 10 numbers are collected

Down below you can see one of the dataset sheets



Extract and classify datasets from these sheets is possible with aruco or corner detection

You can see some of these extracted characters and numbers down below

![pic1_10](https://user-images.githubusercontent.com/47675705/134947355-82918cca-3223-4371-9084-0952665bcb09.png)  ![pic4_7](https://user-images.githubusercontent.com/47675705/134947407-c31aa19e-74f1-4f15-9f9f-55d25964c725.png)  ![pic295_1](https://user-images.githubusercontent.com/47675705/134947605-9592e9f1-eba0-4a00-b00b-9105f575e691.png)  ![pic274_8](https://user-images.githubusercontent.com/47675705/134952721-ea57c389-2c23-4372-b981-d394a6559da2.png)

After training a CNN network with 28x28 input dimention and 42 outputs(persian characters + numbers) by extracted characters I reached about 90 percent of train and 88 percent of test accuracy

Better and wider dataset can help improve accuracy so much