Landslides are one of the most dangerous natural disasters, which result in numerous fatalities, destruction of property and infrastructure across the globe, with thousands of people affected and millions of dollars lost annually. Conventional methods of detection present a major challenge in accurately and efficiently detecting landslides due to the fact that the conventional methods cannot easily capture the complexity of the visuals of the areas prone to landslides. Machine learning and deep learning techniques have been used in the past for detection but they also suffer from a few challenges, such as data insufficiency and generalization across various terrains. Therefore, in this study, a deep learning model based on convolutional neural networks (CNNs) with augmented data and transfer of learning techniques is proposed that includes rotation, scaling, and flipping of data to overcome this challenge, enhance the diversity of the dataset and also to handle the imbalanced class problem, whereas, MobileNetV2, VGG16, ResNet50 and EfficientNetB0 are used as the pre-trained models, after which they are fine-tuned to make them more specific. The evaluation metrics for this hybrid model give better precision, recall, F1 score, IoU and accuracy, which shows that this hybrid model is more useful and stable for large-scale landslide detection and early warning systems.