Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Adaptively Augmented Consistency Learning: A Semi-supervised Segmentation Framework for Remote Sensing

Hui Ye, Haodong Chen, Xiaoming Chen, Vera Chung

TL;DR

AACL is proposed, a semi-supervised segmentation framework designed to enhances RS segmentation accuracy under condictions of limited labeled data and extracts additional information embedded in unlabeled images through the use of Uniform Strength Augmentation and Adaptive Cut-Mix.

Abstract

Remote sensing (RS) involves the acquisition of data about objects or areas from a distance, primarily to monitor environmental changes, manage resources, and support planning and disaster response. A significant challenge in RS segmentation is the scarcity of high-quality labeled images due to the diversity and complexity of RS image, which makes pixel-level annotation difficult and hinders the development of effective supervised segmentation algorithms. To solve this problem, we propose Adaptively Augmented Consistency Learning (AACL), a semi-supervised segmentation framework designed to enhances RS segmentation accuracy under condictions of limited labeled data. AACL extracts additional information embedded in unlabeled images through the use of Uniform Strength Augmentation (USAug) and Adaptive Cut-Mix (AdaCM). Evaluations across various RS datasets demonstrate that AACL achieves competitive performance in semi-supervised segmentation, showing up to a 20% improvement in specific categories and 2% increase in overall performance compared to state-of-the-art frameworks.

Adaptively Augmented Consistency Learning: A Semi-supervised Segmentation Framework for Remote Sensing

TL;DR

AACL is proposed, a semi-supervised segmentation framework designed to enhances RS segmentation accuracy under condictions of limited labeled data and extracts additional information embedded in unlabeled images through the use of Uniform Strength Augmentation and Adaptive Cut-Mix.

Abstract

Remote sensing (RS) involves the acquisition of data about objects or areas from a distance, primarily to monitor environmental changes, manage resources, and support planning and disaster response. A significant challenge in RS segmentation is the scarcity of high-quality labeled images due to the diversity and complexity of RS image, which makes pixel-level annotation difficult and hinders the development of effective supervised segmentation algorithms. To solve this problem, we propose Adaptively Augmented Consistency Learning (AACL), a semi-supervised segmentation framework designed to enhances RS segmentation accuracy under condictions of limited labeled data. AACL extracts additional information embedded in unlabeled images through the use of Uniform Strength Augmentation (USAug) and Adaptive Cut-Mix (AdaCM). Evaluations across various RS datasets demonstrate that AACL achieves competitive performance in semi-supervised segmentation, showing up to a 20% improvement in specific categories and 2% increase in overall performance compared to state-of-the-art frameworks.

Paper Structure

This paper contains 23 sections, 9 equations, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Semi-supervised Image Segmentation
  4. Image Segmentation
  5. Semi-supervised Learning
  6. Semi-supervised Remote Sensing Image Segmentation
  7. Methodology
  8. Problem Definition
  9. Adaptively Augmented Consistency Learning (AACL)
  10. Uniform Strength Augmentation
  11. Adaptive CutMix
  12. Experiment
  13. Experimental Setup
  14. Datasets
  15. Implementation Setting
  16. ...and 8 more sections

Figures (4)

  • Figure 1: Comparison of natural image and remote sensing image. The image on the left side is from Pascal VOC dataset Pascal2015everingham, the image on the right side is from Potsdam dataset.
  • Figure 2: Overview structure of AACL. "$A_w$" and "$A_s$" indicate the weak augmentation and the "USAug" module, respectively.
  • Figure 3: Visualization of strong augmentation applied in USAug. The image on the left is the original representation, the images on the right are the processed images under different strong augmentation. (a) Contrast. (b) Equalize. (c) Blur. (d) Brightness. (e) Saturation. (f) Sharpness. (g) Posterize. (h) Solarize. (i) Hue. (j) Grayscale.
  • Figure 4: The structure of AdaCM module. $I^u$, $I^{u,aux}$ and $I^{l,aux}$ are unlabeled image, auxiliary unlabeled image and auxiliary labeled image, respectively. "$r$" is the random threshold in AdaCM, "$\alpha$" is the trigger probability from the weakly augmented prediction.