Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Proteomic composition and mutual assembly of the C2a projection in vertebrate motile cilia.

eLife·2026
Same author

Molecular Recognition Mechanism of Key VOCs by Odorant-Binding Proteins in the Western Corn Rootworm (<i>Diabrotica virgifera virgifera</i>).

Insects·2026
Same author

Microtubule-associated CCDC112 is essential for spermiogenesis and male fertility in mice.

Journal of molecular cell biology·2026
Same author

RGS22 is a metazoa-specific radial spoke component required for coordinated ciliary beating.

Nature communications·2026
Same author

USP21-mediated deubiquitylation stimulates NuMA recruitment to the cell cortex to promote mitotic spindle orientation.

Cell death and differentiation·2026
Same author

Characterization of membrane structures regulating primary ciliogenesis by quantitative isotropic ultrastructure imaging.

Nature communications·2026

Related Experiment Video

Updated: Feb 28, 2026

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
04:48

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

3.6K

A Systematic Study on Pretraining Strategies for Low-Label Remote Sensing Image Semantic Segmentation.

Peizhuo Liu1, Hongbo Zhu2, Xiaofei Mi2

  • 1School of Instrumentation and Optoelectronic Engineering, Beihang University, Beijing 100191, China.

Sensors (Basel, Switzerland)
|February 27, 2026
PubMed
Summary

This study enhances semantic segmentation for remote sensing images (RSIs) with limited data by optimizing self-supervised pretraining. A novel two-phase approach, General-Purpose Pretraining (GPPT) and Domain-Adaptive Pretraining (DAPT), improves local feature learning for better segmentation accuracy.

Keywords:
limited labelsremote sensingself-supervised learningsemantic segmentationswin transformertwo-stage pretraining

More Related Videos

Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
08:25

Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment

Published on: May 7, 2019

9.7K
Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application
05:56

Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application

Published on: April 14, 2023

3.3K

Related Experiment Videos

Last Updated: Feb 28, 2026

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
04:48

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

3.6K
Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
08:25

Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment

Published on: May 7, 2019

9.7K
Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application
05:56

Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application

Published on: April 14, 2023

3.3K

Area of Science:

  • Computer Vision
  • Remote Sensing
  • Machine Learning

Background:

  • Semantic segmentation of remote sensing images (RSIs) is crucial but challenging with limited labeled data.
  • High-quality initial models are essential for effective semi-supervised and weakly supervised learning in this domain.
  • Existing self-supervised pretraining methods face limitations due to domain shift and poor local feature learning.

Purpose of the Study:

  • To systematically investigate and optimize self-supervised pretraining strategies for RSIs under extreme low-label conditions.
  • To address the domain shift problem and enhance local feature representation in visual foundation models (VFMs).
  • To develop a robust pretraining framework that improves semantic segmentation performance for RSIs.

Main Methods:

  • Benchmarking various pretraining strategies, including single-phase and two-phase approaches.
  • Implementing a two-phase General-Purpose Pretraining (GPPT) followed by Domain-Adaptive Pretraining (DAPT) framework.
  • Proposing an Edge-Guided Masked Image Modeling (EGMIM) method for the DAPT phase to integrate edge priors.

Main Results:

  • The proposed GPPT+DAPT framework significantly outperforms single-phase methods and existing two-phase paradigms.
  • EGMIM effectively enhances the learning of fine-grained local structures by guiding the masking and reconstruction process.
  • Consistent and substantial performance gains were observed across four RSI benchmarks, especially in extreme low-label scenarios.

Conclusions:

  • The optimized two-phase pretraining strategy (GPPT+DAPT) is highly effective for semantic segmentation of RSIs with limited data.
  • Integrating edge information via EGMIM during domain adaptation is crucial for improving local feature learning.
  • The study provides a robust solution for improving RSI semantic segmentation, particularly in data-scarce environments.