In art education, the quality assessment system is crucial; nevertheless, the system’s accuracy is an issue. The ineffectiveness and inadequacy of traditional mechanical learning methods make them unsuitable for use in art education. As a result, this study describes and evaluates a method for assessing the quality of art instruction that makes use of convolutional neural networks. As a first step in reducing interference elements in the quality assessment system, the indicators are split according to the needs of the system, and the influencing variables are located using neural theory. Then, using neuron theory as a basis, we construct the convolution neural network quality assessment system scheme and conduct a thorough analysis of the system’s output. According to the findings of the MATLAB simulations, conventional machine learning is beaten out by convolution neural networks when it comes to the accuracy and speed of the quality assessment system’s influencing elements under specific evaluation criteria.