A Comprehensive Analysis of Techniques and Applications in Multimodal Deep Learning

In recent years, deep learning has gained popularity and been used in many different applications. Models that can process and link data utilizing several modalities are the aim of multimodal deep learning (MMDL). Even with all the advancements made, unimodal learning is still unable to fully capture the entirety of human learning. Multimodal learning facilitates a better understanding and analysis of situations in which information is processed through multiple senses. The various modalities covered in this article include picture, video, text, audio, flow, RGB, stance, depth, face expressions, body motions, and point clouds. There has been a thorough examination of the most current developments in multimodal deep learning applications in the last few years, as well as a detailed investigation of the baseline methodologies. A detailed taxonomy of distinct multimodal deep learning techniques is presented, providing further details on specific applications. Finally, key issues are outlined for each domain separately, along with potential avenues for further research.