Remembering by Reconstructing: Domain Incremental Learning With Test-Time Training on Video Streams

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Swinnen, Jonathan, Tuytelaars, Tinne
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916065008680960
author Swinnen, Jonathan
Tuytelaars, Tinne
author_facet Swinnen, Jonathan
Tuytelaars, Tinne
contents In this work we introduce a novel approach to domain incremental learning, adapting models over time to evolving, non-stationary data. In contrast to other works, we do not attempt to avoid catastrophic forgetting, but rather allow it and exploit it. Our model combines a main task head with a self-supervised masked autoencoder (MAE) head. We then learn domain-specific LoRA adapters during incremental training. Each adapter specializes to its domain, naturally inducing forgetting on other domains in both heads. At inference, we perform online test-time training on the self-supervised MAE head to identify which LoRAs best matches the current input, so the model can `remember' the domain again. Our scheme is especially well-suited to real-world streaming data, such as video, where consecutive samples are highly correlated and domain shifts are gradual. We demonstrate our method on domain-incremental action recognition and semantic segmentation tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2605_31108
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Remembering by Reconstructing: Domain Incremental Learning With Test-Time Training on Video Streams
Swinnen, Jonathan
Tuytelaars, Tinne
Computer Vision and Pattern Recognition
Machine Learning
In this work we introduce a novel approach to domain incremental learning, adapting models over time to evolving, non-stationary data. In contrast to other works, we do not attempt to avoid catastrophic forgetting, but rather allow it and exploit it. Our model combines a main task head with a self-supervised masked autoencoder (MAE) head. We then learn domain-specific LoRA adapters during incremental training. Each adapter specializes to its domain, naturally inducing forgetting on other domains in both heads. At inference, we perform online test-time training on the self-supervised MAE head to identify which LoRAs best matches the current input, so the model can `remember' the domain again. Our scheme is especially well-suited to real-world streaming data, such as video, where consecutive samples are highly correlated and domain shifts are gradual. We demonstrate our method on domain-incremental action recognition and semantic segmentation tasks.
title Remembering by Reconstructing: Domain Incremental Learning With Test-Time Training on Video Streams
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2605.31108