MixerCA: An Efficient and Accurate Model for High-Performance Hyperspectral Image Classification

MixerCA is a lightweight hyperspectral image classification model combining depthwise convolutions, token-channel mixing, and coordinate attention to decouple spatial-spectral interactions while maintaining consistent resolution across the network. Evaluated on four standard HSI benchmarks, it directly processes raw hyperspectral patches without dimensionality reduction preprocessing. Remote sensing engineers and geospatial ML practitioners should prioritize this over heavier transformer-based HSI classifiers where compute efficiency matters.