Saved in:
Bibliographic Details
Main Authors: Sosa, Jose, Aloulou, Mohamed, Rukhovich, Danila, Sleimi, Rim, Changaival, Boonyarit, Kacem, Anis, Aouada, Djamila
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2409.18536
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929517536215040
author Sosa, Jose
Aloulou, Mohamed
Rukhovich, Danila
Sleimi, Rim
Changaival, Boonyarit
Kacem, Anis
Aouada, Djamila
author_facet Sosa, Jose
Aloulou, Mohamed
Rukhovich, Danila
Sleimi, Rim
Changaival, Boonyarit
Kacem, Anis
Aouada, Djamila
contents Self-supervised pre-training has proven highly effective for many computer vision tasks, particularly when labelled data are scarce. In the context of Earth Observation (EO), foundation models and various other Vision Transformer (ViT)-based approaches have been successfully applied for transfer learning to downstream tasks. However, it remains unclear under which conditions pre-trained models offer significant advantages over training from scratch. In this study, we investigate the effectiveness of pre-training ViT-based Masked Autoencoders (MAE) for downstream EO tasks, focusing on reconstruction, segmentation, and classification. We consider two large ViT-based MAE pre-trained models: a foundation model (Prithvi) and SatMAE. We evaluate Prithvi on reconstruction and segmentation-based downstream tasks, and for SatMAE we assess its performance on a classification downstream task. Our findings suggest that pre-training is particularly beneficial when the fine-tuning task closely resembles the pre-training task, e.g. reconstruction. In contrast, for tasks such as segmentation or classification, training from scratch with specific hyperparameter adjustments proved to be equally or more effective.
format Preprint
id arxiv_https___arxiv_org_abs_2409_18536
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle How Effective is Pre-training of Large Masked Autoencoders for Downstream Earth Observation Tasks?
Sosa, Jose
Aloulou, Mohamed
Rukhovich, Danila
Sleimi, Rim
Changaival, Boonyarit
Kacem, Anis
Aouada, Djamila
Computer Vision and Pattern Recognition
Self-supervised pre-training has proven highly effective for many computer vision tasks, particularly when labelled data are scarce. In the context of Earth Observation (EO), foundation models and various other Vision Transformer (ViT)-based approaches have been successfully applied for transfer learning to downstream tasks. However, it remains unclear under which conditions pre-trained models offer significant advantages over training from scratch. In this study, we investigate the effectiveness of pre-training ViT-based Masked Autoencoders (MAE) for downstream EO tasks, focusing on reconstruction, segmentation, and classification. We consider two large ViT-based MAE pre-trained models: a foundation model (Prithvi) and SatMAE. We evaluate Prithvi on reconstruction and segmentation-based downstream tasks, and for SatMAE we assess its performance on a classification downstream task. Our findings suggest that pre-training is particularly beneficial when the fine-tuning task closely resembles the pre-training task, e.g. reconstruction. In contrast, for tasks such as segmentation or classification, training from scratch with specific hyperparameter adjustments proved to be equally or more effective.
title How Effective is Pre-training of Large Masked Autoencoders for Downstream Earth Observation Tasks?
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2409.18536