Turbine blades, which carry approximately one-third of a turbine weight, are susceptible to damage and can cause sudden malfunction of a grid-connected wind energy conversion system. Early detection of blade damage is crucial for preventing catastrophic failures that can lead to downtime, repair costs, and even injury or loss of life. This manuscript aims to explore an image analytics-based deep learning framework for wind turbine blade erosion detection. Turbine blade images are captured via drones/unmanned aerial vehicles during the data collection phase. Upon inspection, it was found that the image dataset was limited; hence, image augmentation was applied to improve the blade image dataset. The approach is modeled as a multiclass supervised learning problem where different turbine blade surface damage scenarios are considered. The potential capability of transfer learning methods such as VGG16-RCNN and AlexNet are tested against a convolutional neural network for detecting the blade’s surface damage. Results reveal the VGG16-RCNN model as the best classifier among the tested ones with the highest accuracy and precision score. To validate the effect of image augmentation on the training data, the accuracy of the proposed VGG16-RCNN framework is assessed via sensitivity analysis, and results of the same reveal that horizontal and vertical flip together with zoom and rotation brings out an efficiency of 93.8%. However, a more generic model that works well with turbines located in different topological regions could be of more importance.