Deep Spectral Network with Vision Transformers for Hyperspectral Image Classification
Working on a Hyperspectral Image (HSI) Classification project using a Dual-Branch architecture called DSNet-ViT, which combines:
A Spectral Unmixing Module (to extract physically meaningful spectral features),
A Vision Transformer (ViT) Module (to capture spatial context and relationships),
And a Fusion Strategy (to merge outputs from both branches for classification).
Enhance classification accuracy by leveraging both spectral purity and spatial structure across benchmark datasets like Indian Pines, Salinas, and Pavia University.