Creating 3D point clouds from 2D images is a fundamental challenge in computer vision, particularly for applications in virtual reality (VR), augmented reality (AR), and robotics. While 2D images provide rich information, they lack depth, making it difficult to understand spatial relationships among objects in a scene. This paper addresses the challenge of reconstructing accurate 3D spatial information from single 2D images, particularly in complex scenes. We propose a method using a convolutional neural network (CNN) for monocular depth estimation, specifically trained to generate high-quality furniture images suitable for use by interior designers. Our model predicts depth maps for each pixel, enabling the reconstruction of a 3D scene as a point cloud. We detail the architecture of our model and the methodology utilized, along with a comprehensive evaluation of experimental results using the NYU Depth V2 dataset. Our findings demonstrate significant improvements in depth accuracy and computational efficiency when compared to existing approaches. Additionally, we discuss the implications of our results for applications in AR, 3D modeling, and robotics, while outlining challenges and proposing future enhancements to further optimize performance in real-world scenarios.