Sentiment analysis on images using different transfer learning models

Abstract

Opinion expression using visual images has grown more common due to the growth of social media platforms and internet technology. Users can communicate their ideas, emotions, and attitudes by posting images to various social media sites and sharing them. In recent years, models based on transfer learning have demonstrated exceptional results for image classification. Successful image-based sentiment analysis is made possible with transfer learning strategies. This study compares and investigates various transfer learning techniques to categorize image sentiment. The findings are analyzed and contrasted using well-known image sentiment datasets, including CK+, FER2013, and JAFFE. Several parameters, including accuracy, precision, and recall, are considered when analyzing the outcomes. The findings indicate that the model's performance depends on the dataset, i.e., different models work well with other datasets.