SquconvNet: Deep Sequencer Convolutional Network for Hyperspectral Image Classification

The application of Transformer in computer vision has had the most significant influence all deep learning developments over past five years. In addition to exceptional performance convolutional neural networks (CNN) hyperspectral image (HSI) classification, begun be applied HSI classification. However, for time being, not produced satisfactory results Recently, field creators Sequencer have proposed a structure that substitutes self-attention layer with BiLSTM2D and achieves results. As result, this paper proposes unique network called SquconvNet, combines CNN block improve paper, we conducted rigorous classification experiments on three relevant baseline datasets evaluate method. experimental show our method clear advantages terms accuracy stability.