Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Hyperspectral Image Classification using Spectral-Spatial Mixer Network

Mohammed Q. Alkhatib

TL;DR

SS-MixNet addresses hyperspectral image classification under limited labeled data by combining a 3D-CNN front-end with two parallel MLP-style mixer blocks for spectral and spatial mixing, supplemented by a lightweight depthwise attention module. It operates on PCA-reduced patches to model long-range spectral and spatial dependencies efficiently, achieving state-of-the-art OA on QUH-Tangdaowan (95.68%) and QUH-Qingyun (93.86%) with only 1% labeled data. Ablation confirms the complementary benefits of the spectral and spatial mixers and the attention mechanism, while maintaining a compact footprint (~141k parameters) and low compute. The approach offers practical, scalable improvements for remote sensing classification in data-constrained settings, with avenues for cross-domain and self-/semi-supervised extensions.

Abstract

This paper introduces SS-MixNet, a lightweight and effective deep learning model for hyperspectral image (HSI) classification. The architecture integrates 3D convolutional layers for local spectral-spatial feature extraction with two parallel MLP-style mixer blocks that capture long-range dependencies in spectral and spatial dimensions. A depthwise convolution-based attention mechanism is employed to enhance discriminative capability with minimal computational overhead. The model is evaluated on the QUH-Tangdaowan and QUH-Qingyun datasets using only 1% of labeled data for training and validation. SS-MixNet achieves the highest performance among compared methods, including 2D-CNN, 3D-CNN, IP-SWIN, SimPoolFormer, and HybridKAN, reaching 95.68% and 93.86% overall accuracy on the Tangdaowan and Qingyun datasets, respectively. The results, supported by quantitative metrics and classification maps, confirm the model's effectiveness in delivering accurate and robust predictions with limited supervision. The code will be made publicly available at: https://github.com/mqalkhatib/SS-MixNet

Hyperspectral Image Classification using Spectral-Spatial Mixer Network

TL;DR

SS-MixNet addresses hyperspectral image classification under limited labeled data by combining a 3D-CNN front-end with two parallel MLP-style mixer blocks for spectral and spatial mixing, supplemented by a lightweight depthwise attention module. It operates on PCA-reduced patches to model long-range spectral and spatial dependencies efficiently, achieving state-of-the-art OA on QUH-Tangdaowan (95.68%) and QUH-Qingyun (93.86%) with only 1% labeled data. Ablation confirms the complementary benefits of the spectral and spatial mixers and the attention mechanism, while maintaining a compact footprint (~141k parameters) and low compute. The approach offers practical, scalable improvements for remote sensing classification in data-constrained settings, with avenues for cross-domain and self-/semi-supervised extensions.

Abstract

This paper introduces SS-MixNet, a lightweight and effective deep learning model for hyperspectral image (HSI) classification. The architecture integrates 3D convolutional layers for local spectral-spatial feature extraction with two parallel MLP-style mixer blocks that capture long-range dependencies in spectral and spatial dimensions. A depthwise convolution-based attention mechanism is employed to enhance discriminative capability with minimal computational overhead. The model is evaluated on the QUH-Tangdaowan and QUH-Qingyun datasets using only 1% of labeled data for training and validation. SS-MixNet achieves the highest performance among compared methods, including 2D-CNN, 3D-CNN, IP-SWIN, SimPoolFormer, and HybridKAN, reaching 95.68% and 93.86% overall accuracy on the Tangdaowan and Qingyun datasets, respectively. The results, supported by quantitative metrics and classification maps, confirm the model's effectiveness in delivering accurate and robust predictions with limited supervision. The code will be made publicly available at: https://github.com/mqalkhatib/SS-MixNet

Paper Structure

This paper contains 8 sections, 2 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. NETWORK ARCHITECTURE
  3. Feature Extraction using 3D-CNN
  4. Spectral Mixer Block
  5. Spatial Mixer Block
  6. Channel-wise Spatial Attention
  7. Experiments and Analysis
  8. Conclusion

Figures (5)

  • Figure 1: (a) Overall architecture of the proposed SS-MixNet Model; (b) Spectral Mixer Block; (c) Spatial Mixer Block.
  • Figure 2: Depthwise Attention Block
  • Figure 3: Reference Data: (a) QUH-Tangdaowan; (b) QUH-Qingyun.
  • Figure 4: Classification maps of Tangdaowan Dataset. (a) 2D-CNN; (b) 3D-CNN; (c) IP-SWIN; (d) SimPoolFormer (e) HybridKAN; (f) Proposed.
  • Figure 5: Classification maps of Qingyun Dataset. (a) 2D-CNN; (b) 3D-CNN; (c) IP-SWIN; (d) SimPoolFormer (e) HybridKAN; (f) Proposed.