<p>Class imbalance is one of the major issues for the application of deep learning for remote sensing imagery. High-resolution remote sensing image sample sets are prone to cause the problem of sample imbalance among classes due to the skewed distribution of ground features, which has been a huge challenge to machine learning and data mining and aroused strong attention. Despite the emerging interest in Deep Learning (DL), empirical research on its effectiveness with imbalanced data in remote sensing remains scarce. Therefore, this study aims to systematically review existing studies using DL approaches for handling class imbalanced data in the field of remote sensing, focusing on its significance in critical applications such as land use land cover classification, pixel-wise segmentation, and object detection. This study presents the most widely used balancing algorithms published from 2016 to 2024. This survey divulges that while using DL technologies, either the data augmentation can be applied to the minority class images to generate more varied samples or using algorithm-level methods by incorporating modifications to the loss function such as cross-entropy, focal loss, dice loss by assigning higher weights to minority class samples, ensuring the model pays more attention to underrepresented classes during training. While traditional techniques such as data sampling and cost-sensitive learning continue to be relevant, emerging methods such as GAN, which leverage neural network feature learning capabilities, show promise. Our discussion identifies research gaps and provides insights to guide future efforts in utilizing DL for imbalanced data in remote sensing applications.</p>