Glaucoma is a disorder of the eyes for which there is no recognized treatment and which, if unchecked, can result in irreversible visual loss. The early and accurate diagnosis of glaucoma, a primary contributor to irreversible blindness, is of paramount importance in ophthalmology. Fundus images have become a valuable resource for aiding in the diagnosis of this condition. In recent years, transfer learning, a powerful approach in the field of deep learning, has gained significant attention for its potential to enhance glaucoma detection accuracy. Transfer learning involves passing the pre-trained model's weights to the new task, where it can be adjusted or used as a feature extractor. This survey paper provides a comprehensive examination of transfer learning approaches employed in glaucoma detection from fundus images. This study compares pre-trained models like VGG16, ResNet, GoogLeNet, MobileNet, DenseNet, and Inception providing insights into their performance, strengths and limitations in glaucoma detection using various datasets and methodologies.