This paper presents a novel approach to hyperspectral image (HSI) classification using deep convolutional neural networks (CNNs). The authors propose a five-layer CNN architecture, including an input layer, a convolutional layer, a max pooling layer, a fully connected layer, and an output layer. This architecture is designed to directly classify HSI data in the spectral domain, leveraging the rich spectral information available in HSI. The experimental results on several HSI datasets, such as Indian Pines, Salinas, and University of Pavia, demonstrate that the proposed method outperforms traditional methods like support vector machines (SVM) and conventional deep learning-based methods. The CNN classifier achieves higher classification accuracy, even with limited training samples, and shows competitive computational costs compared to other neural network architectures. The paper also discusses the training process, including forward and back propagation, and provides detailed comparisons with other neural network architectures. The authors conclude that their CNN-based method is effective for HSI classification and suggests future directions, including the use of Siamese Networks and unsupervised learning to improve performance and reduce the need for labeled samples.This paper presents a novel approach to hyperspectral image (HSI) classification using deep convolutional neural networks (CNNs). The authors propose a five-layer CNN architecture, including an input layer, a convolutional layer, a max pooling layer, a fully connected layer, and an output layer. This architecture is designed to directly classify HSI data in the spectral domain, leveraging the rich spectral information available in HSI. The experimental results on several HSI datasets, such as Indian Pines, Salinas, and University of Pavia, demonstrate that the proposed method outperforms traditional methods like support vector machines (SVM) and conventional deep learning-based methods. The CNN classifier achieves higher classification accuracy, even with limited training samples, and shows competitive computational costs compared to other neural network architectures. The paper also discusses the training process, including forward and back propagation, and provides detailed comparisons with other neural network architectures. The authors conclude that their CNN-based method is effective for HSI classification and suggests future directions, including the use of Siamese Networks and unsupervised learning to improve performance and reduce the need for labeled samples.