 Recent research has demonstrated that using spectral spatial information can significantly improve the accuracy of hyperspectral image, HSI, classification. This is because 3D convolutional neural networks, 3DCNNs, are able to extract both spectral and spatial features from the HSI cube data together, resulting in more accurate classification. Additionally, 3DCNNs require fewer parameters than other deep learning-based methods, making them lighter, less prone to overfitting, and easier to train. When compared to other deep learning-based methods, such as stacked autoencoders, SAEs, deep brief networks, DBNs, and 2DCNN-based methods, the proposed 3DCNN-based method outperformed all of them on three real-world HSI datasets. This article was authored by Eng Lee, Hao Kuai Zhang, and Chong Shen.