This research proposes a deep triplet network for detecting retinal abnormalities from OCT images with limited data. The model incorporates a conditional loss function to enhance accuracy and address overfitting issues. The network is based on the VGG16 architecture and uses a triplet loss function to measure distances between feature embeddings. The model is trained on a public OCT dataset containing 84,000 images, with a subset used for training and validation. The model achieves an overall accuracy of 92.81% in classifying retinal diseases into four categories: Normal, CNV, DME, and Drusen. The proposed model outperforms existing deep learning models in terms of accuracy and generalization, particularly on smaller datasets. The study highlights the effectiveness of the deep triplet network in improving diagnostic precision for retinal disease detection using OCT images.This research proposes a deep triplet network for detecting retinal abnormalities from OCT images with limited data. The model incorporates a conditional loss function to enhance accuracy and address overfitting issues. The network is based on the VGG16 architecture and uses a triplet loss function to measure distances between feature embeddings. The model is trained on a public OCT dataset containing 84,000 images, with a subset used for training and validation. The model achieves an overall accuracy of 92.81% in classifying retinal diseases into four categories: Normal, CNV, DME, and Drusen. The proposed model outperforms existing deep learning models in terms of accuracy and generalization, particularly on smaller datasets. The study highlights the effectiveness of the deep triplet network in improving diagnostic precision for retinal disease detection using OCT images.