A method for generating training data for endoscopic applications, in which stored image data comprising individual images or video data from endoscopic procedures/examinations are prepared as training data for machine learning applications. Wherein the method includes: (a) subjecting the image data to an image recognition algorithm for determining: (i) whether the stored image data was captured inside or outside a body, and (ii) whether text information and faces are contained in the stored image data, and (b) subjecting the image data to an image processing algorithm, with which the image data, depending on the result in (a): (i) are anonymized by obscuring or removing text information and faces when the stored image data is determined to have been captured outside of a body, and (ii) are anonymized by obscuring or removing text information when the stored image data is determined to have been captured inside of a body.