Clustering high-dimensional image data is a critical challenge in fields like computer vision, where the vast amount of information in images requires effective dimensionality reduction techniques to reveal meaningful patterns. In the context of interior design images, grouping similar designs can assist in content-based image retrieval, design recommendations, and automated categorization. However, traditional clustering methods, such as K-Means, struggle with high-dimensional data due to the “curse of dimensionality.” This study addresses this problem by applying two advanced dimensionality reduction techniques-distributed Stochastic Neighbor Embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP)—to extract meaningful lower-dimensional representations of the image data. Feature vectors were extracted from the images using a pre-trained VGG16 model, further reduced using autoencoders, and clustered using K-Means. The performance of the clustering was evaluated using the Silhouette Score, with UMAP achieving a superior score of 0.4149, compared to t-SNE’s score of 0.3729, indicating that UMAP better preserves the global and local structures of the data. The implications of this work are significant for the development of content-based image retrieval systems and automated design recommendation platforms, where well-separated and coherent clusters enhance the relevance and accuracy of recommendations.