This research presents the development of a mobile application designed for the pre-screening of pterygium with the aim of enhancing early detection and improving model accuracy through continuous data collection. Leveraging the power of pre-trained deep learning model approach, the application integrates a smartphone camera with API service for image classification to provide an accessible and user-friendly tool for pterygium detection. The study evaluates several deep learning architectures, including EfficientNetB0, ResNet50, and VGG16, through 5-fold cross-validation and on a separate test set, assessing their precision, recall, and F1 scores. Evaluation classification results and evidence analysis with Grad-CAM function demonstrate that this approach offers a promising rapid solution for enhancing pterygium detection and allows for the data collection of anonymized patient images to continuously refine the model.