In the Thin-Film Transistor Liquid-Crystal Display (TFT-LCD) manufacturing, conducting a machine learning based system with multiple data types has become urgently desired to solve complicated problems. This paper proposes a novel deep learning approach: TabVisionNet, which is modeled by utilizing the information from both tabular data and image data. Tabular data and image data are first encoded by a Tabular encoder and a CNN encoder respectively, then a novel attention mechanism called Sequential Decision Attention was integrated into the framework, that combines the information from two modalities. This novel approach can capture the complex relationship between modalities then gain better generalization and faster convergence in the training process. Conducting an experiment, the performance of our novel approach was significantly better than single-modal and other multi-modal learning approaches in our real case scenario.