This paper presents a computer vision system that recognizes three-dimensional objects from single two-dimensional images without relying on depth reconstruction. The system uses three key mechanisms: perceptual organization to identify image groupings invariant to viewpoint, probabilistic ranking to reduce search space, and spatial correspondence to align 3D models with images by solving for viewpoint and model parameters. The system is robust to occlusion and missing data through viewpoint consistency constraints. It is argued that similar mechanisms underpin human vision. The paper discusses the role of depth reconstruction in human vision, challenging the assumption that it is essential for object recognition. Human vision can recognize images without depth cues, such as simple line drawings, and uses partial and ambiguous information to achieve reliable identification. The paper argues that depth reconstruction is not the primary pathway for recognition in human vision. The system, SCERPO, uses spatial correspondence to match 3D models with images by solving for viewpoint and model parameters. It uses Newton's method to iteratively refine these parameters, allowing for accurate verification of matches. The system is robust to missing data and can handle complex backgrounds. The paper also discusses the use of line-to-line correspondences and parameter determination for matching. It presents methods for extending initial matches and using probabilistic approaches to select reliable matches. The system is designed to be robust against missing data and occlusion, ensuring accurate recognition even with incomplete information. The paper concludes that the system's ability to recognize objects from 2D images without depth reconstruction is supported by the principles of perceptual organization and viewpoint invariance. These principles allow the system to detect stable image groupings that reflect actual scene structure, even in the presence of ambiguity and noise. The system's robustness and efficiency make it a promising approach for computer vision applications.This paper presents a computer vision system that recognizes three-dimensional objects from single two-dimensional images without relying on depth reconstruction. The system uses three key mechanisms: perceptual organization to identify image groupings invariant to viewpoint, probabilistic ranking to reduce search space, and spatial correspondence to align 3D models with images by solving for viewpoint and model parameters. The system is robust to occlusion and missing data through viewpoint consistency constraints. It is argued that similar mechanisms underpin human vision. The paper discusses the role of depth reconstruction in human vision, challenging the assumption that it is essential for object recognition. Human vision can recognize images without depth cues, such as simple line drawings, and uses partial and ambiguous information to achieve reliable identification. The paper argues that depth reconstruction is not the primary pathway for recognition in human vision. The system, SCERPO, uses spatial correspondence to match 3D models with images by solving for viewpoint and model parameters. It uses Newton's method to iteratively refine these parameters, allowing for accurate verification of matches. The system is robust to missing data and can handle complex backgrounds. The paper also discusses the use of line-to-line correspondences and parameter determination for matching. It presents methods for extending initial matches and using probabilistic approaches to select reliable matches. The system is designed to be robust against missing data and occlusion, ensuring accurate recognition even with incomplete information. The paper concludes that the system's ability to recognize objects from 2D images without depth reconstruction is supported by the principles of perceptual organization and viewpoint invariance. These principles allow the system to detect stable image groupings that reflect actual scene structure, even in the presence of ambiguity and noise. The system's robustness and efficiency make it a promising approach for computer vision applications.