An End-to-End Offline Handwritten Tamil Text Identification Using Modified RAdam Optimizer with Effective Post-processing Techniques

The foremost aim of the recommended, automatic identification of handwritten Tamil text is offline identification of the manually written Tamil script in an editable display. The novel system used to obtain the best accuracy rate with less processing time by the transfer learning skill with the Inception-v3 deep learning model. This suggested research plan uses the modified RAdam optimizer, which optimizes the entire model in a better way to achieve stable training and model generalization. An attempt initiated to improve the recognition accuracy with the help of spelling error detection and correction of the recognized Tamil text in the post-processing stage. Additionally, the identified manually written Tamil content converted into the English language with an altered Google translation framework. Experimental results on the Tamil handwritten text identification verify the perception and prove the effectiveness and robustness of the modified RAdam optimizer. The trial outcomes achieved from this proposed research work reveal the efficiency of the automatic Tamil script identification system in an enhanced height.