Awesome
45K+ Clean-Background-Urdu-Ligatures-Dataset
Urdu Ligatures dataset / 38K-Training images & 7K-Testing images / Uniformly imbedded Ligatures
<b>Synthetic ->> Urdu Ligatures Dataset for Urdu Text Recognition </b>
<p> <img src="SampleImages/855_Original_RetuSinu.jpg" width=200> <img src="SampleImages/36_Original_RetuNoisy1.jpg" width=200> <img src="SampleImages/2068_Original_RetuBarrel.jpg" width=200> <img src="SampleImages/5_Original_RetuNoisy12.jpg" width=200> <img src="SampleImages/2074_Original_RetuSinu.jpg" width=200> <img src="SampleImages/27_Original_RetuNRSI.jpg" width=200> <img src="SampleImages/0_Original_RetuNoisy12.jpg" width=200> <img src="SampleImages/2483_Original_RetuNoisy12.jpg" width=200> </p> # Urdu Ligatures Dataset (Synthetic 45K+ images)--->>> The <b>Given dataset</b> here is <b>most similar to dataset</b> we used in our exmperiments/research with minor differences. The images and their annotation are given in text with its filenames contain Urdu ligature information and name of files representing types of transformations applied. <b>So, from given annotation information and images they can be used for training different machine learning algorithms. .<br> .<br>
Cleaned Background based Urdu Ligatures
--->>> The <b>Urdu-Ligatures-Dataset</b> consisting of more than 45000 images (also corresponding 45000 annotations) is <b>most similar to the dataset</b> we used in our exmperiments/research with minor differences. Images are given in zipped format and the <b>annotations are in text files</b> .<br> .<br> <br> !<br> !<br> !<br> !<br> !<br> Please cite our dataset as!! <br> <b> Syed Yasser Arafat, and Muhmmad Javed Iqbal. "Two Stream Deep Neural Network for Sequence-Based Urdu Ligature Recognition." IEEE Access 7 (2019): 159090-159099.. </b> <br> <br> <br> <br>
<div> <p> for any queries: <b > [Syed Yasser Arafat]: <mailto:syed.yasser.arafat@gmail.com> </b> </p> </div> syed.yasser.arafat@gmail.com