Uncorrected refractive errors are the leading cause of preventable vision impairment globally, disproportionately affecting individuals in low-resource regions. Despite the availability of economical treatments, challenges in timely diagnosis and access to screening persist, particularly in underserved communities. This study introduces a novel deep learning-based system for automated refractive error classification using photorefractive images acquired via a standard smartphone camera. A multi-branch convolutional neural network (CNN) was developed and trained on a dataset of 2,139 corneal images from an Indonesian sample, enabling the model to classify refractive errors into four categories: significant myopia, significant hypermetropia, insignificant refractive error, and not applicable to classified. The 3-branch CNN architecture demonstrated superior performance, achieving an overall test accuracy of 91%, precision of 96%, and recall of 98%, with an area under the curve (AUC) score of 0.9896. Its multi-scale feature extraction pathways were pivotal in addressing overlapping red reflex patterns and subtle variations between classes. The dataset, collected from a public eye hospital in Indonesia, reflects real-world diversity and strengthens the model's reliability for deployment in resource-limited settings. Grad-CAM visualisation provided insights into the model's interpretability, enhancing its clinical applicability. This study establishes the feasibility of smartphone-based photorefractive assessment integrated with artificial intelligence for scalable, cost-effective vision screening. By training the CNN model with data representative of Southeast Asian populations, this system offers a reliable solution for early refractive error detection, with significant implications for improving accessibility to eye care services. © 2025 Copyright held by the owner/author(s).