Deep Pyramidal Residual Networks for Spectral–Spatial Hyperspectral Image Classification

By Mercedes Eugenia Paoletti, Juan Mario Haut, Rubén Fernández-Beltran, Javier Plaza, Antonio J. Plaza, Filiberto Pla. IEEE Transactions on Geoscience and Remote Sensing, 2019

428 citations recorded by the source corpus.

Convolutional neural networks (CNNs) exhibit good performance in image processing tasks, pointing themselves as the current state-of-the-art of deep learning methods. However, the intrinsic complexity of remotely sensed hyperspectral images still limits the performance of many CNN models. The high dimensionality of the HSI data, together with the underlying redundancy and noise, often makes the standard CNN approaches unable to generalize discriminative spectral–spatial features. Moreover, deeper CNN architectures also find challenges when additional layers are added, which hampers the network convergence and produces low classification accuracies. In order to mitigate these issues, this paper presents a new deep CNN architecture specially designed for the HSI data. Our new model pursues to improve the spectral–spatial features uncovered by the convolutional filters of the network. Specifically, the proposed residual-based approach gradually increases the feature map dimension at all convolutional layers, grouped in pyramidal bottleneck residual blocks, in order to involve more locations as the network depth increases while balancing the workload among all units, preserving the time complexity per layer. It can be seen as a pyramid, where the deeper the blocks, the more feature maps can be extracted. Therefore, the diversity of high-level spectral–spatial attributes can be gradually increased across layers to enhance the performance of the proposed network with the HSI data. Our experiments, conducted using four well-known HSI data sets and 10 different classification techniques, reveal that our newly developed HSI pyramidal residual model is able to provide competitive advantages (in terms of both classification accuracy and computational time) over the state-of-the-art HSI classification methods

Subfields: Atmospheric Science, Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/1701daaccd6a5a5ff882b64dd5534988788e32f5

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Atmospheric Science

In Computer Vision and Pattern Recognition

In Media Technology

Citations in the corpus

Spectral–Spatial Morphological Attention Transformer for Hyperspectral Image Classification (2023, background): “To take full advantage of ResNets, they can be extended to form even more complex models, such as the inclusion of adaptive kernels [17], lightweight spatial–spectral attention based on squeeze-and-excitation [35], and pyramidal ResNets [37].”
Capsule Networks for Hyperspectral Image Classification (2019, background): “In this sense, the improvements brought to CNN filters (kernels) via the development of residual connections [33]–[35] (ResNet) and dense skip connections [36], [37] (DenseNet) open new and interesting paths to uncover highly discriminative spectral–spatial features present in HSI data.”
Attention-Based Adaptive Spectral–Spatial Kernel ResNet for Hyperspectral Image Classification (2020, methodology): “The poor performance is given by ResNet and DPyResNet in terms of OA and kappa on all data sets.”
Multimodal Fusion Transformer for Remote Sensing Image Classification (2022, background): “To take advantage of the residual behavior of the network, one can extend it to form an even more complex model, i.e., the lightweight spectral-spatial squeeze-and-excitation attention [38] with adaptive kernels [20] and pyramidal residual networks [41].”
Hyperspectral Image Classification—Traditional to Deep Models: A Survey for Future Prospects (2021, background): “Recently, joint exploitation of both spectral and spatial features has gained much popularity and led to improved classification accuracy [49], [115]–[118].”
Deep Learning for Hyperspectral Image Classification: An Overview (2019, background): “In addition, the works in [87], [93], [97]–[99] introduced residual learning (RL) [100]”
Spectral–Spatial Feature Tokenization Transformer for Hyperspectral Image Classification (2022, methodology): “In [43], a pyramid residual network was developed to classify HSI.”
HybridSN: Exploring 3-D–2-D CNN Feature Hierarchy for Hyperspectral Image Classification (2019, background): “have proposed the Deep Pyramidal Residual Networks (DPRN) specially for the HSI data [29].”
A survey: Deep learning for hyperspectral image classification with few labeled samples (2021, background): “Moreover, due to the success of ResNet [59], scholars have studied the HSI classification problem based on convolutional residuals [57, 58, 60, 61].”