Fig. 4From: CLoDSA: a tool for augmentation in classification, localization, detection, semantic segmentation and instance segmentation tasksProcess to automatically label augmented images for the semantic segmentation problem. From the original image (top left) and the annotation image (bottom left), two new images are generated by applying the transformation (in this case a 90∘ rotation) to both of them (top right and bottom right images). Images obtained from [20], these images are available under a Attribution-NonCommercial 3.0 Unported licenceBack to article page