This paper presents a 3D Convolutional Neural Network (3D-CNN) framework for the classification of hyperspectral images (HSIs). The proposed method leverages the 3D structure of HSI data, which is typically presented as 3D cubes, to simultaneously extract spectral and spatial features. Unlike traditional methods that rely on preprocessing or post-processing, 3D-CNN directly processes the 3D cube data, reducing the need for dimensionality reduction and preserving spatial information. The model is designed to be lightweight, with fewer parameters compared to other deep learning-based methods, making it less prone to overfitting and easier to train. The effectiveness of the 3D-CNN is evaluated on three real-world HSI datasets from different sensors, and the results show that the proposed method outperforms state-of-the-art methods, including stacked autoencoders (SAE), deep belief networks (DBN), and 2D-CNN-based approaches. The 3D-CNN achieves higher overall accuracy, average accuracy, and kappa statistics, demonstrating its superior performance in HSI classification. The paper also discusses the impact of various parameters on the model's performance and suggests future research directions, such as integrating unsupervised and semi-supervised classification methods to better utilize unlabeled samples.This paper presents a 3D Convolutional Neural Network (3D-CNN) framework for the classification of hyperspectral images (HSIs). The proposed method leverages the 3D structure of HSI data, which is typically presented as 3D cubes, to simultaneously extract spectral and spatial features. Unlike traditional methods that rely on preprocessing or post-processing, 3D-CNN directly processes the 3D cube data, reducing the need for dimensionality reduction and preserving spatial information. The model is designed to be lightweight, with fewer parameters compared to other deep learning-based methods, making it less prone to overfitting and easier to train. The effectiveness of the 3D-CNN is evaluated on three real-world HSI datasets from different sensors, and the results show that the proposed method outperforms state-of-the-art methods, including stacked autoencoders (SAE), deep belief networks (DBN), and 2D-CNN-based approaches. The 3D-CNN achieves higher overall accuracy, average accuracy, and kappa statistics, demonstrating its superior performance in HSI classification. The paper also discusses the impact of various parameters on the model's performance and suggests future research directions, such as integrating unsupervised and semi-supervised classification methods to better utilize unlabeled samples.