The remarkable growth of deep learning (DL) algorithms in hyperspectral images (HSIs) in recent years has garnered a lot of research space. This study examines and analyses over 250 newly proposed breakthroughs, the majority from 2019 onwards, that focussed on DL methods for HSI classification. Initially, a meta-analysis was performed to analyse the state-of-the-art literature, followed by an in-depth investigation of the performance of DL models on benchmark hyperspectral datasets. It provides an adequate grasp of how DL models are employed with the suggested improvisations and evaluates the performance of these state-of-the-art models. It investigates the accuracy of benchmark hyperspectral datasets by applying the newly proposed DL models. This study discusses the challenges that researchers have faced and the solutions that have been proposed. In particular, the scarcity of non-traditional hyperspectral data sets with labels impedes the performance of novel DL algorithms. As a result, this article clarifies the supervised, semi-supervised, and unsupervised training modes employed by the models and analyses the performance of the DL models while utilizing them.

Convolutional neural networks have demonstrated remarkable performance in capturing discriminative features in the hyperspectral domain. Deep learning and CNN have carved out a distinct niche in the remote-sensing community having the potential to classify high-dimensional images. However, the challenge of limited labelled samples and high dimensional hyperspectral data aggravates overfitting in deep networks. Many existing works produce high accuracy but are computationally slower due to the massive computations involved. The Inception model, on the other hand, won the ImageNet Large Scale Visual Recognition Competition (ILSVRC) 2014 challenge and has been validated as a top-tier model by maintaining a trade-off between classification accuracy and computing time. In this paper, we propose a novel lightweight network, Xcep-Dense, a hybrid classification model that combines the core benefits of the extreme version of inception and dense networks. The Xception network employs depth-wise separable convolutions, and the 3D slicing phenomenon, which requires fewer parameters, is computationally efficient and provides excellent classification accuracy. The proposed network is configured with dense network and optimization parameters to alleviate overfitting and gradient vanishing. Xcep-Dense’s performance is validated using two benchmark hyperspectral datasets, Indian Pines and Salinas.