The data for deep learning is growing bigger day by day. That leads to traditional processing systems facing computing resource limitations. So, distributed parallel computing system for training deep learing models is needed. Within the scope of this paprer, we deploy some traditional deep learning models such as Lenet and Alexnet for Vietnamese food image classification. Model performance is relatively satisfactory, with an accuracy of 49% with Lenet. Details of the data processing and model architecture are explained in the sections below.