Skip to Main content Skip to Navigation
Conference papers

Continual Learning for Dense Labeling of Satellite Images

Onur Tasar 1 Yuliya Tarabalka 1 Pierre Alliez 1
1 TITANE - Geometric Modeling of 3D Environments
CRISAM - Inria Sophia Antipolis - Méditerranée
Abstract : n dense labeling problem, the major drawback of the convolutional neural networks is their inability to learn new classes without affecting performance for the old classes on the data, having no annotations for the previous classes. In this work, we address the issue of adding new classes continually to the already trained network from a stream of data. Our approach comprises two main components: adaptation and remembering. For adaptation, we keep a clone of the previously trained network, which serves as a memory for the old classes in absence of their annotations on the new data. The updated network learns new as well as old classes on the current data using output of the memory network and the new groundtruth. For remembering, we store a little portion of the previous data, from which we systematically feed samples to the updated network during training. Our results prove that segmentation capabilities for the new classes can be added to the already trained network without catastrophically forgetting the previously learned information.
Complete list of metadatas

Cited literature [11 references]  Display  Hide  Download

https://hal.inria.fr/hal-02276543
Contributor : Onur Tasar <>
Submitted on : Monday, September 2, 2019 - 5:24:10 PM
Last modification on : Monday, October 19, 2020 - 3:28:45 AM
Long-term archiving on: : Thursday, January 9, 2020 - 4:18:28 PM

File

igarss19.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-02276543, version 1

Citation

Onur Tasar, Yuliya Tarabalka, Pierre Alliez. Continual Learning for Dense Labeling of Satellite Images. IGARSS 2019 - IEEE International Geoscience and Remote Sensing Symposium, Jul 2019, Yokohama, Japan. ⟨hal-02276543⟩

Share

Metrics

Record views

96

Files downloads

447