Knowledge Distillation for Incremental Learning in Semantic Segmentation

TL;DR

This paper introduces a novel approach for incremental learning in semantic segmentation using knowledge distillation techniques to prevent catastrophic forgetting, enabling models to learn new classes without losing previous knowledge.

Contribution

It is the first to apply and adapt knowledge distillation for incremental learning specifically in semantic segmentation tasks, with four new methodologies developed.

Findings

Effective retention of previous class knowledge

High accuracy maintained on old classes during incremental learning

Successful application on Pascal VOC2012 and MSRC-v2 datasets

Abstract

Deep learning architectures have shown remarkable results in scene understanding problems, however they exhibit a critical drop of performances when they are required to learn incrementally new tasks without forgetting old ones. This catastrophic forgetting phenomenon impacts on the deployment of artificial intelligence in real world scenarios where systems need to learn new and different representations over time. Current approaches for incremental learning deal only with image classification and object detection tasks, while in this work we formally introduce incremental learning for semantic segmentation. We tackle the problem applying various knowledge distillation techniques on the previous model. In this way, we retain the information about learned classes, whilst updating the current model to learn the new ones. We developed four main methodologies of knowledge distillation working on both output layers and internal feature representations. We do not store any image belonging to previous training stages and only the last model is used to preserve high accuracy on previously learned classes. Extensive experimental results on the Pascal VOC2012 and MSRC-v2 datasets show the effectiveness of the proposed approaches in several incremental learning scenarios.