Boosting Semi-Supervised Scene Text Recognition via Viewing and Summarizing