<p>The deficiency of perceptual demands and the inefficiency of product development have been persistent issues in the mechanical product manufacturing sector. The reasons lie in the absence of an effective standardized process during the mechanical product design to integrate the characteristics of mechanical products, corporate objectives, and user requirements. Meanwhile, there is a notable scarcity of effective strategies to mitigate the constraints imposed by mechanical product design on the work efficiency of designers. To address the problems in the form design process of mechanical products mentioned above, this paper proposes a three-dimensional (3D) form generation and perceptual evaluation process based on prototype theory, topology, and deep learning, and verifies its effectiveness and practicality in real-world applications. This process covers multiple models throughout the mechanical product form design process: the form and style mapping model, which utilizes the correlation among performance, imagery, and form to construct a 3D form that meets user perceptual demands and reflects product performance; the form perceptual recognition model and generation model, which consider the style constraint characteristics in the mechanical product design process and use deep learning and form quantification techniques to quickly complete the fine-grained style imagery recognition and directional style generation of 3D forms. The presented research on tractor form design demonstrates that this design process can quickly obtain product forms that conform to user style imagery and reflect enterprise product performance, and showcases the specific process of this design method, thereby providing a universal and effective reference for mechanical product form design. At the same time, we also point out the maturity of this process and the work needed to improve its usability in the future.</p>