Early Preparation Pays Off: New Classifier Pre-tuning for Class Incremental Semantic Segmentation

TL;DR

This paper introduces NeST, a pre-tuning method that transforms old classifiers to initialize new ones, improving class incremental semantic segmentation by reducing catastrophic forgetting and background shift.

Contribution

NeST is a novel pre-tuning approach that generates new classifiers from old ones before training, enhancing stability and adaptability in incremental learning.

Findings

Significant performance improvements on Pascal VOC 2012 and ADE20K datasets.

Effective alignment of new classifiers with the backbone and data.

Enhanced stability-plasticity trade-off through class similarity-based initialization.

Abstract

Class incremental semantic segmentation aims to preserve old knowledge while learning new tasks, however, it is impeded by catastrophic forgetting and background shift issues. Prior works indicate the pivotal importance of initializing new classifiers and mainly focus on transferring knowledge from the background classifier or preparing classifiers for future classes, neglecting the flexibility and variance of new classifiers. In this paper, we propose a new classifier pre-tuning~(NeST) method applied before the formal training process, learning a transformation from old classifiers to generate new classifiers for initialization rather than directly tuning the parameters of new classifiers. Our method can make new classifiers align with the backbone and adapt to the new data, preventing drastic changes in the feature extractor when learning new classes. Besides, we design a strategy considering the cross-task class similarity to initialize matrices used in the transformation, helping achieve the stability-plasticity trade-off. Experiments on Pascal VOC 2012 and ADE20K datasets show that the proposed strategy can significantly improve the performance of previous methods. The code is available at \url{https://github.com/zhengyuan-xie/ECCV24_NeST}.