Improving the Accuracy of Prediction of Plant Diseases Using Dimensionality Reduction-Based Ensemble Models
01 Jan 2020-pp 121-129
TL;DR: The proposed research work aims to improve the performance of classification tasks on diseased plants by exploring t-distributed Stochastic Neighbor Embedding (t-SNE) based Ensemble Models.
Abstract: In many real-world applications, different features can be obtained and how to duly utilize them in reduced dimension is a challenge. Simply concatenating them into a long vector is not appropriate because each view has its specific statistical property and physical interpretation. Many dimensionality reduction methods have been developed to identify this lower-dimensional space and map data to it, reducing the number of predictors in supervised learning problems and allowing for better visualization of data relations and clusters. However, the plethora of dimensionality reduction techniques provides a variety of nonlinear, linear, global, and local methods, and it is likely that each method captures different data features. Ensemble methods have achieved much success in supervised learning, from Random Forest to AdaBoost. Ensembles exploit diversity and balance bias, variance, and covariance to achieve these results is likely that disparate dimensionality reduction methods will enhance diversity within a dimensionality reduction-based ensemble. AdaBoost and Random Forest are popular ensemble methods which are widely used for classification of target variables. Major problem with ensembles like AdaBoost and Random Forest is that they perform worse when dimensionality of data is high. Random Forest is the predictor ensemble with a set of decision trees that grow in randomly selected subspaces of data. The proposed research work aims to improve the performance of classification tasks on diseased plants by exploring t-distributed Stochastic Neighbor Embedding (t-SNE) based Ensemble Models. The infected and healthy plant images are subjected to deep learning model to produce their corresponding image embedding. The high dimensional data with thousands of features is then reduced to a smaller number of features dataset by the state-of-the-art t-SNE algorithm. The significant feature dataset is then given as input to the ensemble models to perform the prediction task.
Citations
More filters
01 Jul 2021
TL;DR: This study presents a new deep learning approach based on capsule networks and ensemble learning for the detection of plant diseases using multi-channel capsule networks, individually trained on images applied different preprocessing techniques and then combined together.
Abstract: This study presents a new deep learning approach based on capsule networks and ensemble learning for the detection of plant diseases. The developed method is called as multi-channel capsule network ensemble. The main innovation behind the proposed method is the use of multi-channel capsule networks, individually trained on images applied different preprocessing techniques and then combined together. In this way, the final ensemble can better detect plant diseases by making use of different attributes of the data. Our experiments carried out using a well-known data set and various state-of-the-art classification methods demonstrate that our classification approach can provide competitive advantages in terms of classification accuracy.
8 citations
TL;DR: In this paper , the authors compared the accuracy rate and root mean square error (RMSE) of the prediction system by using temperature data to predict crop disease, and found that the Max Voting method performed better than the KNN method.
4 citations
TL;DR: In this paper , a graph-based neural network embedding approach was proposed to extract appropriate features into latent space and circumvent the spectrums' nonlinearity problem by constructing a nearest neighbor graph and producing almost linear embedding using a fully connected neural network.
Abstract: Spectroscopy is a methodology for gaining knowledge of particles, especially biomolecules, by quantifying the interactions between matter and light. By examining the level of light absorbed, reflected or released by a specimen, its constituents, properties, and volume can be determined. Spectra obtained through spectroscopy procedures are quick, harmless and contactless; hence nowadays preferred in chemometrics. Due to the high dimensional nature of the spectra, it is challenging to build a robust classifier with good performance metrics. Many linear and nonlinear dimensionality reduction-based classification models have been previously implemented to overcome this issue. However, they lack in capturing the subtle details of the spectra into the low dimension space or cannot efficiently handle the nonlinearity present in the spectral data. We propose a graph-based neural network embedding approach to extract appropriate features into latent space and circumvent the spectrums' nonlinearity problem. Our approach performs dimensionality reduction into two phases: constructing a nearest neighbor graph and producing almost linear embedding using a fully connected neural network. Further, the low dimensional embedding is subjected to classification using the Random Forest algorithm. In this paper, we have implemented and compared our technique with four nonlinear dimensionality techniques widely used for spectral data analysis. In this study, we have considered five different spectral datasets belonging to specific applications. The various classification performance metrics of all the techniques are evaluated. The proposed approach is able to perform competitively well on six different low-dimensional spaces for each dataset with an accuracy score above 95% and Matthew's correlation coefficient value close to 1. The trustworthiness score of almost 1 show that the presented dimensionality reduction approach preserves the closest neighbor structure of high dimensional spectral inputs into latent space.
References
More filters
Proceedings Article•
04 Sep 2014TL;DR: This work investigates the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting using an architecture with very small convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers.
Abstract: In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.
55,235 citations
Proceedings Article•
01 Jan 2015TL;DR: In this paper, the authors investigated the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting and showed that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 layers.
Abstract: In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.
49,914 citations
Journal Article•
TL;DR: A new technique called t-SNE that visualizes high-dimensional data by giving each datapoint a location in a two or three-dimensional map, a variation of Stochastic Neighbor Embedding that is much easier to optimize, and produces significantly better visualizations by reducing the tendency to crowd points together in the center of the map.
Abstract: We present a new technique called “t-SNE” that visualizes high-dimensional data by giving each datapoint a location in a two or three-dimensional map. The technique is a variation of Stochastic Neighbor Embedding (Hinton and Roweis, 2002) that is much easier to optimize, and produces significantly better visualizations by reducing the tendency to crowd points together in the center of the map. t-SNE is better than existing techniques at creating a single map that reveals structure at many different scales. This is particularly important for high-dimensional data that lie on several different, but related, low-dimensional manifolds, such as images of objects from multiple classes seen from multiple viewpoints. For visualizing the structure of very large datasets, we show how t-SNE can use random walks on neighborhood graphs to allow the implicit structure of all of the data to influence the way in which a subset of the data is displayed. We illustrate the performance of t-SNE on a wide variety of datasets and compare it with many other non-parametric visualization techniques, including Sammon mapping, Isomap, and Locally Linear Embedding. The visualizations produced by t-SNE are significantly better than those produced by the other techniques on almost all of the datasets.
30,124 citations
01 Sep 2018
TL;DR: A Convolutional Neural Network model and Learning Vector Quantization algorithm based method for tomato leaf disease detection and classification and results validate that the proposed method effectively recognizes four different types of tomato leaf diseases.
Abstract: The early detection of diseases is important in agriculture for an efficient crop yield. The bacterial spot, late blight, septoria leaf spot and yellow curved leaf diseases affect the crop quality of tomatoes. Automatic methods for classification of plant diseases also help taking action after detecting the symptoms of leaf diseases. This paper presents a Convolutional Neural Network (CNN) model and Learning Vector Quantization (LVQ) algorithm based method for tomato leaf disease detection and classification. The dataset contains 500 images of tomato leaves with four symptoms of diseases. We have modeled a CNN for automatic feature extraction and classification. Color information is actively used for plant leaf disease researches. In our model, the filters are applied to three channels based on RGB components. The LVQ has been fed with the output feature vector of convolution part for training the network. The experimental results validate that the proposed method effectively recognizes four different types of tomato leaf diseases.
265 citations
TL;DR: A method named as bacterial foraging optimization based radial basis function neural network (BRBFNN) for identification and classification of plant leaf diseases automatically and attains higher accuracy in identification and Classification of diseases is introduced.
Abstract: The contribution of a plant is highly important for both human life and environment. Plants do suffer from diseases, like human beings and animals. There is the number of plant diseases that occur and affects the normal growth of a plant. These diseases affect complete plant including leaf, stem, fruit, root, and flower. Most of the time when the disease of a plant has not been taken care of, the plant dies or may cause leaves drop, flowers, and fruits drop. Appropriate diagnosis of such diseases is required for accurate identification and treatment of plant diseases. Plant pathology is the study of plant diseases, their causes, procedures for controlling and managing them. But, the existing method encompasses human involvement for classification and identification of diseases. This procedure is time-consuming and costly. Automatic segmentation of diseases from plant leaf images using soft computing approach can be reasonably useful than the existing one. In this paper, we have introduced a method named as bacterial foraging optimization based radial basis function neural network (BRBFNN) for identification and classification of plant leaf diseases automatically. For assigning optimal weight to radial basis function neural network we use bacterial foraging optimization that further increases the speed and accuracy of the network to identify and classify the regions infected of different diseases on the plant leafs. The region growing algorithm increases the efficiency of the network by searching and grouping of seed points having common attributes for feature extraction process. We worked on fungal diseases like common rust, cedar apple rust, late blight, leaf curl, leaf spot, and early blight. The proposed method attains higher accuracy in identification and classification of diseases.
171 citations