Guided contextual attention mapping for repair tasks

Systems and methods for augmenting data may utilize one or more machine-learned models and contextual attention data to provide more realistic and efficient data augmentation. For example, systems and methods for repair may utilize a machine-learned model to generate predicted contextual attention data and mix the predicted contextual attention data with the obtained contextual attention data to determine replacement data for enhancing an image to replace one or more occlusions. The obtained contextual attention data may include user-guided contextual attention. 用于增强数据的系统和方法可以利用一个或多个机器学习的模型和上下文注意力数据来提供更真实和有效的数据增强。例如，用于修复的系统和方法可以利用机器学习的模型来生成预测的上下文注意力数据，并将预测的上下文注意力数据与获得的上下文注意力数据混合，以确定用于增强图像以替换一个或多个遮挡的替换数据。获得的上下文注意力数据可以包括用户引导的上下文注意力。