RECALL: Replay-based Continual Learning in Semantic Segmentation

TL;DR

RECALL introduces a replay-based continual learning method for semantic segmentation that generates or retrieves past class data without storing samples, effectively mitigating catastrophic forgetting during incremental learning.

Contribution

It proposes a novel replay strategy using GAN-generated and web-crawled data to prevent forgetting without storing previous samples.

Findings

RECALL outperforms existing methods in incremental semantic segmentation.

The approach effectively mitigates catastrophic forgetting.

No storage of past data is required, preserving privacy.

Abstract

Deep networks allow to obtain outstanding results in semantic segmentation, however they need to be trained in a single shot with a large amount of data. Continual learning settings where new classes are learned in incremental steps and previous training data is no longer available are challenging due to the catastrophic forgetting phenomenon. Existing approaches typically fail when several incremental steps are performed or in presence of a distribution shift of the background class. We tackle these issues by recreating no longer available data for the old classes and outlining a content inpainting scheme on the background class. We propose two sources for replay data. The first resorts to a generative adversarial network to sample from the class space of past learning steps. The second relies on web-crawled data to retrieve images containing examples of old classes from online databases. In both scenarios no samples of past steps are stored, thus avoiding privacy concerns. Replay data are then blended with new samples during the incremental steps. Our approach, RECALL, outperforms state-of-the-art methods.