 This paper proposes an end-to-end fast-dent spectral end-dash. Spatial convolution, F keys SC, framework for hyperspectral image, HSI, classification that uses different convolutional kernel sizes to extract spectral and spatial features separately, densely connected structures for deep learning of features, dynamic learning rate, parametric rectified linear units, batch normalization, and dropout layers to achieve accuracy within as few as 80 epochs while significantly reducing the training time. The experimental results show that the proposed FDSSC framework achieves state-of-the-art performance compared with existing deep learning-based methods on Indian pines, Kennedy Space Center, and University of Povea data sets. This article was authored by Wenju Wang, Shibuang Do, Zhongming Jiang, and others.