Image Text Translation Technology based on Time Convolutional Network Transformer Model

Text translation is a process Machine translation from image to text by computer programs to translate natural source language into a precise other language. However, translation text from image has drawback of ineffective deep learning models which translate the words inaccurate due to missing important keyword extraction from the image. To overcome these limitations, a Time Convolutional Network based Transformer (TCN-Transformer) is proposed to make the text translation process more effective in image to text translation technology. The integration of TCN with transformer is highly adaptive, and analyze the patterns in a student's language use to adapt lessons to their specific learning. By effectively capturing both temporal patterns and contextual meaning, the system can deliver more precise and personalized feedback on pronunciation, grammar, and usage. Initially, the text with image dataset ae acquired and preprocessed by binarization technique and segmented by broken character approach for overlapping characters to improve text translation. Then, significant features from the segmented characters are finally classified and translated by the proposed TCN-Transformer method. The experimental results of the proposed method achieved accuracy of 84.82% which is better than existing method such as Bidirectional Long Short-Term Memory (Bi-LSTM).