An Improved Convolutional Neural Network with LSTM Approach for Texture Classification

Subba Rao M
2020 International Journal of Emerging Trends in Engineering Research  
Texture classification is a problem that has several applications, such as remote detection and recognition of forest species. Solutions tend to be customized for the dataset used, but are not generalized. Machine learning algorithms play an important role in the current Texture Classification. However, these algorithms are suffering with low accuracy and classification rate. Deep learning is another sophisticated technique to solve these challenges because texture classification performance is
more » ... not strong in traditional machine learning systems. The Convolutional Neural Network (CNN) in combination with Long Short-Term Memory (LSTM) forms a robust selection between a powerful invariant feature extractor and an accurate classifier. This model should automatically determine the efficient properties of the feature samples so that the texture samples can be classified accurately. Expert fusion provides stability in classification rates between different data sets and the proposed model will significantly increases texture classification performance. From the experimental analysis, it is ascertained that CNN-LSTM is outperforms with existing state of the art of the algorithms SVM and CNN.
doi:10.30534/ijeter/2020/148872020 fatcat:7szrrlcd7be77fnbntsvezoodm