Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MSCloudCAM: Multi-Scale Context Adaptation with Convolutional Cross-Attention for Multispectral Cloud Segmentation

Md Abdullah Al Mazid, Liangdong Deng, Naphtali Rishe

TL;DR

Clouds significantly hinder optical remote sensing analysis, especially across heterogeneous multispectral sensors. MSCloudCAM introduces a convolution-based cross-attention framework that fuses dual multi-scale context encoders—one prioritizing fine spatial details and the other global semantics—via a learned cross-attention mechanism. The approach, integrated into a Swin Transformer backbone with a deep-supervision decoder, achieves state-of-the-art performance on CloudSEN12 and L8Biome while maintaining competitive complexity. This method provides robust spectral–spatial discrimination across sensors, enabling more reliable cloud masks for environmental and climate analyses.

Abstract

Clouds remain a major obstacle in optical satellite imaging, limiting accurate environmental and climate analysis. To address the strong spectral variability and the large scale differences among cloud types, we propose MSCloudCAM, a novel multi-scale context adapter network with convolution based cross-attention tailored for multispectral and multi-sensor cloud segmentation. A key contribution of MSCloudCAM is the explicit modeling of multiple complementary multi-scale context extractors. And also, rather than simply stacking or concatenating their outputs, our formulation uses one extractor's fine-resolution features and the other extractor's global contextual representations enabling dynamic, scale-aware feature selection. Building on this idea, we design a new convolution-based cross attention adapter that effectively fuses localized, detailed information with broader multi-scale context. Integrated with a hierarchical vision backbone and refined through channel and spatial attention mechanisms, MSCloudCAM achieves strong spectral-spatial discrimination. Experiments on various multisensor datatsets e.g. CloudSEN12 (Sentinel-2) and L8Biome (Landsat-8) show that MSCloudCAM outperforms recent state-of-the-art models while maintaining competitive model complexity, highlighting the novelty and effectiveness of the proposed design for large-scale Earth observation.

MSCloudCAM: Multi-Scale Context Adaptation with Convolutional Cross-Attention for Multispectral Cloud Segmentation

TL;DR

Clouds significantly hinder optical remote sensing analysis, especially across heterogeneous multispectral sensors. MSCloudCAM introduces a convolution-based cross-attention framework that fuses dual multi-scale context encoders—one prioritizing fine spatial details and the other global semantics—via a learned cross-attention mechanism. The approach, integrated into a Swin Transformer backbone with a deep-supervision decoder, achieves state-of-the-art performance on CloudSEN12 and L8Biome while maintaining competitive complexity. This method provides robust spectral–spatial discrimination across sensors, enabling more reliable cloud masks for environmental and climate analyses.

Abstract

Clouds remain a major obstacle in optical satellite imaging, limiting accurate environmental and climate analysis. To address the strong spectral variability and the large scale differences among cloud types, we propose MSCloudCAM, a novel multi-scale context adapter network with convolution based cross-attention tailored for multispectral and multi-sensor cloud segmentation. A key contribution of MSCloudCAM is the explicit modeling of multiple complementary multi-scale context extractors. And also, rather than simply stacking or concatenating their outputs, our formulation uses one extractor's fine-resolution features and the other extractor's global contextual representations enabling dynamic, scale-aware feature selection. Building on this idea, we design a new convolution-based cross attention adapter that effectively fuses localized, detailed information with broader multi-scale context. Integrated with a hierarchical vision backbone and refined through channel and spatial attention mechanisms, MSCloudCAM achieves strong spectral-spatial discrimination. Experiments on various multisensor datatsets e.g. CloudSEN12 (Sentinel-2) and L8Biome (Landsat-8) show that MSCloudCAM outperforms recent state-of-the-art models while maintaining competitive model complexity, highlighting the novelty and effectiveness of the proposed design for large-scale Earth observation.

Paper Structure

This paper contains 25 sections, 11 equations, 3 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Cloud Segmentation in Remote Sensing
  4. Deep Learning Approaches
  5. Gaps in Current Methods
  6. Methodology
  7. Overview
  8. Backbone Feature Extraction
  9. Multi-Scale Context Modules
  10. Cross-Attention Fusion
  11. Bottleneck Projection
  12. Combined Attention Refinement
  13. Decoder with Deep Supervision
  14. Experiments
  15. Datasets
  16. ...and 10 more sections

Figures (3)

  • Figure 1: High-level architecture of MSCloudCAM. The model integrates a hierarchical vision backbone with dual multi-scale context extractors and introduces a novel cross-attention module.
  • Figure 2: Architecture of MSCloudCAM. A multispectral Swin Transformer encoder generates hierarchical features. ASPP and PSP modules enrich deep and intermediate representations. These are fused by convolutional cross-attention, refined via channel–spatial attention, and decoded by a multi-stage upsampling path with auxiliary supervision.
  • Figure 3: Qualitative cloud segmentation results. Each row corresponds to a dataset. The proposed method produces sharper delineation of thin clouds and shadows compared to other approaches. Class–color mapping: Class 1 (Clear sky – light blue), Class 2 (Thick cloud – white), Class 3 (Thin cloud – light gray), Class 4 (Cloud shadow – dark gray).