Dynamic Mode Decomposition Based Multi-View Learning

In computer vision, the classification of 3D objects has garnered significant interest, traditionally tackled using deep learning methods. Despite their effectiveness, these methods come with high computational costs and complexity. This study proposes a novel methodology that combines Dynamic Mode Decomposition (DMD) with preliminary data pre-processing to efficiently extract features and dynamics from multi-angle images of 3D objects. Integrating machine learning classifiers with this framework not only reduces computational demands but also improves the efficiency of multi-view classification. Consequently, this methodology expands its applicability, presenting a viable alternative to the resource-heavy deep learning approaches common in the field. Our research highlights the promising potential of merging DMD with machine learning to enhance 3D object classification in computer vision, marking an important direction for future research.