Recently, Deep learning architectures based on convolutional neural networks have established outstanding performance on different image processing applications such as object recognition, medical image processing, image enhancement, image segmentation, etc. It has shown gigantic improvement in the performance of 2-D and 3-D image processing because of its discriminant and highly connected feature representation capability. This article presents a lightweight deep convolutional neural network (DCNN) framework for hyperspectral image classification in the spectral domain. The proposed DCNN architecture consists of the chief six layers such as the input layer, the convolution layer, the Rectified Linear Unit Layer, the maximum pooling layer, the fully connected layer, and the classifier layer. Experimental results on the Indian-Pines hyperspectral image dataset reveal that the proposed lightweight DCNN can accomplish superior classification performance (98.20% accuracy) than several conventional techniques and the traditional deep learning-based techniques.