Incremental Learning Techniques for Semantic Segmentation
This work addresses incremental learning for semantic segmentation, a domain-specific problem, but it is incremental as it adapts existing distillation techniques to this task.
The paper tackles catastrophic forgetting in deep learning for semantic segmentation by proposing knowledge distillation techniques that transfer information from previous models to new ones without storing old images. The methods achieved high accuracy on the Pascal VOC2012 dataset, demonstrating effectiveness in retaining performance on previously learned classes.
Deep learning architectures exhibit a critical drop of performance due to catastrophic forgetting when they are required to incrementally learn new tasks. Contemporary incremental learning frameworks focus on image classification and object detection while in this work we formally introduce the incremental learning problem for semantic segmentation in which a pixel-wise labeling is considered. To tackle this task we propose to distill the knowledge of the previous model to retain the information about previously learned classes, whilst updating the current model to learn the new ones. We propose various approaches working both on the output logits and on intermediate features. In opposition to some recent frameworks, we do not store any image from previously learned classes and only the last model is needed to preserve high accuracy on these classes. The experimental evaluation on the Pascal VOC2012 dataset shows the effectiveness of the proposed approaches.