Training deep network models for accurate recognition of texts in scene images

Abstract

Recognition of text automatically is playing an important role and act as a foundation in Artificial Intelligence field. In the previous decade, researchers are struggle on overcoming the complicity in their pipeline. With applying deep learning in text recognition, the overall performance and accuracy improved greatly. In this FYP, the state of art deep learning models for text recognition, CRNN and ASTER, will be implemented and trained. For optimal performance, multiple hyperparameter will be tuned. During the chapter of methodology, the issues people might face will be discussed and ways for solving the issues will be provided. The model performance on various datasets will be evaluated and showed in this report.