Age-related macular degeneration (AMD) is a significant cause of vision impairment in older adults. While it doesn’t lead to total blindness, the loss of central vision can make everyday tasks like recognizing faces, reading, driving, and detailed work much more challenging. The progression of AMD varies; for some, it advances gradually, while in others, it can worsen more quickly. In its early stages, AMD may not noticeably affect vision, which is why regular eye exams are essential for early detection and management. Identifying AMD early is key to preserving vision. Deep learning (DL) has emerged as a powerful tool for detecting early lesions and monitoring disease progression in retinal images. It can objectively identify structural changes, stage diseases, and locate specific retinal lesions. In this work, we present a deep transfer learning based MobileNetV3 framework to evaluate the AMD progression through Optical Coherence Tomography. To further improve the performance, we applied different test-time data augmentation techniques which showed significantly better performance for disease evolution and disease progression. We also explored various pretrained models, including DenseNet201, ResNet-34, and DinoV2. However, MobileNetV3 combined with test-time augmentation techniques delivered superior performance, particularly in Task 2, and performed well in Task 1. Our proposed approach achieved F1 scores of 0.727 for Task 2 and 0.716 for Task 1 on the initial leaderboard test set. On the external test dataset, the model’s final overall ranking showed an F1 score of 0.769 for Task 1 and 0.217 for Task 2.