Incremental learning represents a crucial task in aerial image processing, especially given the limited availability of large-scale annotated datasets. A major issue concerning current deep neural architectures is known as catastrophic forgetting, namely the inability to faithfully maintain past knowledge once a new set of data is provided for retraining. Over the years, several techniques have been proposed to mitigate this problem for image classification and object detection. However, only recently the focus has shifted towards more complex downstream tasks such as instance or semantic segmentation. Starting from incremental-class learning for semantic segmentation tasks, our goal is to adapt this strategy to the aerial domain, exploiting a peculiar feature that differentiates it from natural images, namely the orientation. In addition to the standard knowledge distillation approach, we propose a contrastive regularization, where any given input is compared with its augmented version (i.e. flipping and rotations) in order to minimize the difference between the segmentation features produced by both inputs. We show the effectiveness of our solution on the Potsdam dataset, outperforming the incremental baseline in every test.
A Contrastive Distillation Approach for Incremental Semantic Segmentation in Aerial Images / Arnaudo, Edoardo; Cermelli, Fabio; Tavera, Antonio; Rossi, Claudio; Caputo, Barbara. - ELETTRONICO. - (2022), pp. 742-754. (Intervento presentato al convegno International Conference on Image Analysis and Processing tenutosi a Lecce (IT) nel 23/05/2022) [10.1007/978-3-031-06430-2_62].
A Contrastive Distillation Approach for Incremental Semantic Segmentation in Aerial Images
Arnaudo,Edoardo;Cermelli,Fabio;Tavera,Antonio;Caputo,Barbara
2022
Abstract
Incremental learning represents a crucial task in aerial image processing, especially given the limited availability of large-scale annotated datasets. A major issue concerning current deep neural architectures is known as catastrophic forgetting, namely the inability to faithfully maintain past knowledge once a new set of data is provided for retraining. Over the years, several techniques have been proposed to mitigate this problem for image classification and object detection. However, only recently the focus has shifted towards more complex downstream tasks such as instance or semantic segmentation. Starting from incremental-class learning for semantic segmentation tasks, our goal is to adapt this strategy to the aerial domain, exploiting a peculiar feature that differentiates it from natural images, namely the orientation. In addition to the standard knowledge distillation approach, we propose a contrastive regularization, where any given input is compared with its augmented version (i.e. flipping and rotations) in order to minimize the difference between the segmentation features produced by both inputs. We show the effectiveness of our solution on the Potsdam dataset, outperforming the incremental baseline in every test.File | Dimensione | Formato | |
---|---|---|---|
ICIAP_2021_arnaudo_contrastive_distillation_camera_ready.pdf
Open Access dal 13/05/2023
Descrizione: Articolo principale
Tipologia:
2. Post-print / Author's Accepted Manuscript
Licenza:
Pubblico - Tutti i diritti riservati
Dimensione
2.23 MB
Formato
Adobe PDF
|
2.23 MB | Adobe PDF | Visualizza/Apri |
978-3-031-06430-2_62.pdf
accesso riservato
Tipologia:
2a Post-print versione editoriale / Version of Record
Licenza:
Non Pubblico - Accesso privato/ristretto
Dimensione
1.81 MB
Formato
Adobe PDF
|
1.81 MB | Adobe PDF | Visualizza/Apri Richiedi una copia |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/11583/2962571