Urdu Handwritten Text Recognition Using ResNet18
This addresses the need for Urdu HCR systems, which are minimal compared to other languages, but the approach is incremental as it applies an existing method to a new dataset.
The paper tackles Urdu handwritten text recognition, a challenging task due to cursive characters, by proposing a ResNet18 model and achieves results on the UNHD dataset containing 312,000 words.
Handwritten text recognition is an active research area in the field of deep learning and artificial intelligence to convert handwritten text into machine-understandable. A lot of work has been done for other languages, especially for English, but work for the Urdu language is very minimal due to the cursive nature of Urdu characters. The need for Urdu HCR systems is increasing because of the advancement of technology. In this paper, we propose a ResNet18 model for handwritten text recognition using Urdu Nastaliq Handwritten Dataset (UNHD) which contains 3,12000 words written by 500 candidates.