Smoothing Continual Segmentation Oscillations with Latent Domain PPCA Decoder

Marie-Ange Boum, Pierre Fournier, Dawa Derksen, Stéphane Herbin
Proceedings of The TerraBytes {ICML} Workshop: Towards global datasets and models for Earth Observation, PMLR 292:124-140, 2025.

Abstract

We study Domain Incremental Learning for the semantic segmentation of Earth Observation images. We demonstrate that controlling the oscillation of performance when a new domain arrives is more critical than controlling catastrophic forgetting. We propose an exemplar free architecture that combines a large pre-trained network well adapted to dense image processing (DINOv2) and a generative decoder head based on Probabilitic Principal Component Analysis (PPCA). We validate our approach on the FLAIR#1 high resolution dataset, which is structured as a sequence of domains.

Cite this Paper

BibTeX
@InProceedings{pmlr-v292-boum25a, title = {Smoothing Continual Segmentation Oscillations with Latent Domain {PPCA} Decoder}, author = {Boum, Marie-Ange and Fournier, Pierre and Derksen, Dawa and Herbin, St{\'e}phane}, booktitle = {Proceedings of The TerraBytes {ICML} Workshop: Towards global datasets and models for Earth Observation}, pages = {124--140}, year = {2025}, editor = {Audebert, Nicolas and Azizpour, Hossein and Barrière, Valentin and Castillo Navarro, Javiera and Czerkawski, Mikolaj and Fang, Heng and Francis, Alistair and Marsocci, Valerio and Nascetti, Andrea and Yadav, Ritu}, volume = {292}, series = {Proceedings of Machine Learning Research}, month = {19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v292/main/assets/boum25a/boum25a.pdf}, url = {https://proceedings.mlr.press/v292/boum25a.html}, abstract = {We study Domain Incremental Learning for the semantic segmentation of Earth Observation images. We demonstrate that controlling the oscillation of performance when a new domain arrives is more critical than controlling catastrophic forgetting. We propose an exemplar free architecture that combines a large pre-trained network well adapted to dense image processing (DINOv2) and a generative decoder head based on Probabilitic Principal Component Analysis (PPCA). We validate our approach on the FLAIR#1 high resolution dataset, which is structured as a sequence of domains.} }
Endnote
%0 Conference Paper %T Smoothing Continual Segmentation Oscillations with Latent Domain PPCA Decoder %A Marie-Ange Boum %A Pierre Fournier %A Dawa Derksen %A Stéphane Herbin %B Proceedings of The TerraBytes {ICML} Workshop: Towards global datasets and models for Earth Observation %C Proceedings of Machine Learning Research %D 2025 %E Nicolas Audebert %E Hossein Azizpour %E Valentin Barrière %E Javiera Castillo Navarro %E Mikolaj Czerkawski %E Heng Fang %E Alistair Francis %E Valerio Marsocci %E Andrea Nascetti %E Ritu Yadav %F pmlr-v292-boum25a %I PMLR %P 124--140 %U https://proceedings.mlr.press/v292/boum25a.html %V 292 %X We study Domain Incremental Learning for the semantic segmentation of Earth Observation images. We demonstrate that controlling the oscillation of performance when a new domain arrives is more critical than controlling catastrophic forgetting. We propose an exemplar free architecture that combines a large pre-trained network well adapted to dense image processing (DINOv2) and a generative decoder head based on Probabilitic Principal Component Analysis (PPCA). We validate our approach on the FLAIR#1 high resolution dataset, which is structured as a sequence of domains.
APA
Boum, M., Fournier, P., Derksen, D. & Herbin, S.. (2025). Smoothing Continual Segmentation Oscillations with Latent Domain PPCA Decoder. Proceedings of The TerraBytes {ICML} Workshop: Towards global datasets and models for Earth Observation, in Proceedings of Machine Learning Research 292:124-140 Available from https://proceedings.mlr.press/v292/boum25a.html.

Related Material