Hyperspectral Image Classification based on Cycle GAN and EfficientNet

Because of the existence of narrow contiguous bands, Hyperspectral Remote Sensing is a sophisticated remote sensing technique that improves the capacity of precise categorization. Given the availability of duplicate information, the massive quantity of continuous bands observed in hyperspectral photos raises the processing complexity. The primary data sets have a limited number of labeled pixels; however, we need additional pixels for improved hyperspectral picture categorization. As a result, the main hurdles are dimensionality reduction and a restricted number of labelled samples. To address these issues, we introduced the Kernel PCA approach for dimensionality reduction, the Cycle GAN technique for increasing labelled samples, and the EfficientNet technique for classification. These strategies will be more accurate than existing techniques for hyperspectral picture categorization.