<p>Stereoscopy is used in virtual reality to enhance realism and depth perception. Studies about stereoscopic 3D (S3D) images usually focus on evaluating image qualities or assessing visual comfort. The current industry standards or product guidelines related to stereoscopic viewing have not specifically addressed users’ eye conditions such as myopia, hyperopia, astigmatism, asthenopia, strabismus, or previous eye surgery. Different eye conditions and characteristics of human visual system have been proven to have impacts in visual comfort of viewing S3D images. However, information on viewers’ eye conditions is difficult to access since even viewers themselves may not be aware of their eye conditions. In addition, it is impractical and unreasonable to examine users’ eye conditions as in an ophthalmology clinic. This study provided a novel solution to predicting users’ eye conditions with only a few S3D images. These images were selected from a dataset of S3D maps, and each image corresponded to one or several image features, which had been proved to relate to visual discomfort. A clustering model was proposed to cluster viewers’ eye conditions according to their ratings on these images. The performance of the clustering model was evaluated based on viewers’ self-reported eye conditions. This study brings up a new perspective of considering individuals’ eye conditions in the design of S3D content and proves a feasible solution. The future study calls for S3D datasets with large populations and in-depth cooperation among ophthalmologists, human factors professionals, designers, and developers to further investigate the problem and provide standards for guiding the design.</p>