Improving Handwritten Mathematical Expression Recognition via Integrating Convolutional Neural Network With Transformer and Diffusion-Based Data Augmentation
Handwritten mathematical expression recognition (HMER) poses a formidable challenge due to the intricate two-dimensional structures and diverse handwriting styles.This paper introduces a novel approach to improve HMER accuracy by employing an integrated, high-capacity architecture that combines Transformer and Convolutional Neural Network (CNN) mod