Hyperspectral Image Classification and Reconstruction

The field of hyperspectral image classification and reconstruction is moving towards the development of more efficient and effective models that can handle the challenges of high spectral dimensionality, complex spectral-spatial correlations, and limited training samples. Researchers are exploring the use of hybrid architectures that combine the strengths of convolutional neural networks and transformers to improve classification accuracy. Additionally, semi-supervised domain adaptation techniques are being investigated to bridge the gap between general and human-centered hyperspectral image datasets, which is crucial for medical applications. Another area of focus is the development of lightweight and effective deep learning models that can capture long-range dependencies in spectral and spatial dimensions. Noteworthy papers include CLAReSNet, which proposes a hybrid architecture that integrates multi-scale convolutional extraction with transformer-style attention, and SpectralAdapt, which introduces a semi-supervised domain adaptation framework that bridges the domain gap between general and human-centered HSI datasets. SpectralTrain is also a notable framework that enhances learning efficiency by integrating curriculum learning with principal component analysis-based spectral downsampling.

Sources

CLAReSNet: When Convolution Meets Latent Attention for Hyperspectral Image Classification

SpectralAdapt: Semi-Supervised Domain Adaptation with Spectral Priors for Human-Centered Hyperspectral Image Reconstruction

Hyperspectral Image Classification using Spectral-Spatial Mixer Network

SpectralTrain: A Universal Framework for Hyperspectral Image Classification

Built with on top of