# Principles of Forgetting in Domain-Incremental Semantic Segmentation in Adverse Weather Conditions

Source: [Principles of Forgetting in Domain-Incremental Semantic Segmentation in Adverse Weather Conditions](<https://www.sota2.com/research/paper/68fa6fbf1e209131cea47329>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-02T02:58:13.589Z

Authors: Tobias Kalb, J\\"urgen Beyerer

Published: 2023-03-24T16:23:49.000Z

## Abstract

Deep neural networks for scene perception in automated vehicles achieve excellent results for the domains they were trained on. However, in real-world conditions, the domain of operation and its underlying data distribution are subject to change. Adverse weather conditions, in particular, can significantly decrease model performance when such data are not available during training.Additionally, when a model is incrementally adapted to a new domain, it suffers from catastrophic forgetting, causing a significant drop in performance on previously observed domains. Despite recent progress in reducing catastrophic forgetting, its causes and effects remain obscure. Therefore, we study how the representations of semantic segmentation models are affected during domain-incremental learning in adverse weather conditions. Our experiments and representational analyses indicate that catastrophic forgetting is primarily caused by changes to low-level features in domain-incremental learning and that learning more general features on the source domain using pre-training and image augmentations leads to efficient feature reuse in subsequent tasks, which drastically reduces catastrophic forgetting. These findings highlight the importance of methods that facilitate generalized features for effective continual learning algorithms.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Semantic Segmentation on ACDC Night (test)](<https://www.sota2.com/research/sota/semantic-segmentation-on-acdc-night-test.md>): Semantic segmentation; ACDC Night (test); Accuracy: 62.4.

- [Semantic Segmentation on ACDC Rain (test)](<https://www.sota2.com/research/sota/semantic-segmentation-on-acdc-rain-test.md>): Semantic segmentation; ACDC Rain (test); Learning Accuracy: 71.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2303.14115>)

[PDF](<https://arxiv.org/pdf/2303.14115.pdf>)

[Implementation](<https://www.sota2.com/github.com/tobiaskalb/feature-reuse-css>)