Deep neural networks for texture classification—A theoretical analysis

Basu, Supratim; Mukhopadhyay, Supratik; Manohar, Krishpersad; DiBiano, Robert; Ganguly, Sangram; Nemani, Ramakrishna R.; Gayaka, Shreekant

doi:10.1016/j.neunet.2017.10.001

Cited by 85 publications

(31 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…We experimentally show that our framework surpasses all existing stateof-the-art algorithms for high-resolution satellite imagery classification on both SAT-4 and SAT-6 datasets, including the original DeepSAT (Basu et al 2015a), MLP (Zscore) , SatCNN (both Z -score and linear) , TradCNN (Z -score) , triplet networks (Liu and Huang 2018), D-DSML-Caffenet , and contrastive loss (Simo-Serra et al 2015). It has been shown theoretically in (Basu et al 2018(Basu et al , 2016 CNNs, by themselves, are not able to learn representations of Haralick features from data. By augmenting CNNs with the handcrafted features, we are enhancing the discriminative power of CNNs for satellite imagery.…”

Section: Introductionmentioning

confidence: 69%

DeepSat V2: feature augmented convolutional neural nets for satellite image classification

Basu

Ganguly

et al. 2019

Remote Sensing Letters

Self Cite

View full text Add to dashboard Cite

Satellite image classification is a challenging problem that lies at the crossroads of remote sensing, computer vision, and machine learning. Due to the high variability inherent in satellite data, most of the current object classification approaches are not suitable for handling satellite datasets. The progress of satellite image analytics has also been inhibited by the lack of a single labeled high-resolution dataset with multiple class labels.In a preliminary version of this work, we introduced two new high resolution satellite imagery datasets (SAT-4 and SAT-6) and proposed DeepSat framework for classification based on "handcrafted" features and a deep belief network (DBN). The present paper is an extended version, we present an end-to-end framework leveraging an improved architecture that augments a convolutional neural network (CNN) with handcrafted features (instead of using DBN-based architecture) for classification.Our framework, having access to fused spatial information obtained from handcrafted features as well as CNN feature maps, have achieved accuracies of 99.90% and 99.84% respectively, on SAT-4 and SAT-6, surpassing all the other state-ofthe-art results. A statistical analysis based on Distribution Separability Criterion substantiates the robustness of our approach in learning better representations for satellite imagery.

show abstract

Section: Introductionmentioning

confidence: 69%

DeepSat V2: feature augmented convolutional neural nets for satellite image classification

Basu

Ganguly

et al. 2019

Remote Sensing Letters

Self Cite

View full text Add to dashboard Cite

show abstract

“…Other authors provided either a theoretical analysis or visualizing analysis in a context of an application. For example, Basu et al [12] published a theoretical analysis for texture classification whilst Minematsu et al [134,135] provided a visualizing analysis for background subtraction. Despite these first valuable investigation, the understanding of DNNs remains still shallows.…”

Section: Theoretical Aspectsmentioning

confidence: 99%

Deep neural network concepts for background subtraction:A systematic review and comparative evaluation

et al. 2019

View full text Add to dashboard Cite

Conventional neural networks show a powerful framework for background subtraction in video acquired by static cameras. Indeed, the well-known SOBS method and its variants based on neural networks were the leader methods on the largescale CDnet 2012 dataset during a long time. Recently, convolutional neural networks which belong to deep learning methods were employed with success for background initialization, foreground detection and deep learned features. Currently, the top current background subtraction methods in CDnet 2014 are based on deep neural networks with a large gap of performance in comparison on the conventional unsupervised approaches based on multi-features or multi-cues strategies. Furthermore, a huge amount of papers was published since 2016 when Braham and Van Droogenbroeck published their first work on CNN applied to background subtraction providing a regular gain of performance. In this context, we provide the first review of deep neural network concepts in background subtraction for novices and experts in order to analyze this success and to provide further directions. For this, we first surveyed the methods used background initialization, background subtraction and deep learned features. Then, we discuss the adequacy of deep neural networks for background subtraction. Finally, experimental results are presented on the CDnet 2014 dataset.

show abstract

“…The common methods which are used for texture classification are namely parametric statistical model-based methods, structural methods, empirical second order statistical methods and various other transform methods. Deep learningbased techniques for the classification of texture have been proposed in [3][4][5][6].…”

Section: Texture Classificationmentioning

confidence: 99%

Texture classification using convolutional neural network optimized with whale optimization algorithm

et al. 2019

View full text Add to dashboard Cite

Texture classification is an active area of research in the field of pattern recognition. Convolutional neural networks (CNNs) have a remarkable capability of recognizing patterns and are one of the most efficient deep learning techniques. But, finding the optimal values of the different hyperparameters of the CNN is a major challenge. Nature-inspired algorithms (NIAs) are the meta-heuristic algorithms well-known for their optimizing capability. Whale optimization algorithm (WOA) is a recent nature-inspired algorithm (NIA) that is inspired by the hunting behaviour of the humpback whales. In this paper, we propose a novel deep learning technique for texture recognition using a CNN optimized through WOA. We apply WOA at the two different levels in the CNN: In the convolutional layer (for optimizing the values of the filters), and in the fully-connected layer (for optimizing the values of the weights and biases). For examining the performance of our technique, we apply it to the following three benchmark texture datasets: Kylberg v1.0, Brodatz, and Outex_TC_00012. Our model performs better than the most of the existing methods for the Kylberg and the Outex_TC_00012 datasets and gives competitive results for the Brodatz dataset. It is evident from the results that our model has the potential for application in the field of texture recognition.

show abstract

Deep neural networks for texture classification—A theoretical analysis

Cited by 85 publications

References 10 publications

DeepSat V2: feature augmented convolutional neural nets for satellite image classification

DeepSat V2: feature augmented convolutional neural nets for satellite image classification

Deep neural network concepts for background subtraction:A systematic review and comparative evaluation

Texture classification using convolutional neural network optimized with whale optimization algorithm

Contact Info

Product

Resources

About