Multimodal Analysis of Face-Text-Based Promotional Posters

This paper proposed a multimodal approach to assist in analyzing information from promotional posters. Based on advanced Optical Character Recognition (OCR) for text and Generative Adversarial Network (GAN) for facial data it can extract effectively poster content. In addition, the FaceNet's facial recognition ensures the correct identification of an individual. From the experiments, it achieved accuracy rates of 92% for facial recognition, and nearly 99% for text recognition. A simple frontend user interface was developed to improve user experience. These results contribute to further validation in the field of social network information for ensuring security for public information.