A fuzzy LSTM framework for the application of Chinese art painting style classification

Guó Huà is the native painting style of Chinese fine arts, distinguished by its asymmetrical compositions of times and flowers. The painting sketch differs based on the number of strokes, the ages, and the brush used. Therefore, to classify Chinese paintings based on modern computer-based intelligent algorithms, this article introduces a novel method. The proposed method combines fuzzy logic and long short-term memory (LSTM) learning to enhance precision in classifications. The input is a drawing image from which the intensity features are extracted. Using the intensity features, the fuzzy generates a maximum combination of stroke patterns relating to the paintings. A sequence of previously labeled painting features is used to perform pattern matching, utilizing intensity features to reduce dissimilar combinations. Based on the maximum similar combinations, the low and high-intensity vanishing sketch features are detected using the LSTM algorithm. This algorithm is recurrently and individually trained to categorize the sketch’s intensities, lowering or increasing, to differentiate its unique patterns from those of other paintings. The fuzzy derivatives for maximum pattern combinations are converged using the maximum classification precision.