This training explores machine learning methods for enhancing document classification by leveraging both text and image-based content. Traditional methods often emphasize textual elements, neglecting the informative value embedded in visuals. Our work examines how the fusion of these modalities can strengthen classification accuracy, especially in complex documents with varied formats or handwritten annotations. We assess several machine learning algorithms and apply multiple fusion strategies to unify textual and visual data. Evaluations on benchmark datasets reveal substantial performance improvements when using integrated features, demonstrating real-world relevance in document organization and analytics. Future directions include deeper integration using advanced neural models and analysis of diverse visual attributes.