Spontaneous Macro and Micro Facial Expression Recognition Using ResNet50 and VLDSP

Facial expressions serve as the predominant means of conveying human emotions, and their understanding has garnered significant interest among researchers due to the wide array of real-world applications it encompasses. However, there is limited research conducted on the simultaneous recognition of spontaneous macro expressions and micro expressions. This study introduces an innovative end-to-end framework designed to effectively process videos for spontaneous macro and micro facial expression recognition. The framework integrates the extraction of deep spatial features through ResNet50, and handcrafted spatiotemporal features obtained from the application of Volume Local Directional Structural Pattern (VLDSP). Additionally, a novel 1D Convolutional Neural Network (CNN) is introduced to classify facial expressions, utilizing the extracted deep spatial features and spatiotemporal features as input. Finally, a comprehensive experiment is conducted to demonstrate the proposed framework performance.