Automated detection of selected tea leaf diseases in Bangladesh with convolutional neural network

Sendjasni A., Traparic D., Larabi M.

Ogundokun R.O., Maskeliunas R., Misra S., Damaševičius R.

After evaluating the difficulty of CNNs in extracting convolution features, this paper suggested an improved convolutional neural network (CNN) method (ICNN-BNDOA), which is based on Batch Normalization (BN), Dropout (DO), and Adaptive Moment Estimation (Adam) optimizer. To circumvent the gradient challenge and quicken convergence, the ICNN-BNDOA uses a sequential CNN structure with the Leaky rectified linear unit (LeakyReLU) as the activation function (AF). The approach employs an Adam optimizer to handle the overfitting problem, which is done by introducing BN and DO layers to the entire connected CNN layers and the output layers, respectively, to decrease cross-entropy. Through a small regularization impact, BN was utilized to substantially speed up the training process of a neural network, as well as to increase the model's performance. The performance of the proposed system with conventional CNN (CCNN) was studied using the CIFAR-10 datasets as the benchmark data, and it was discovered that the suggested method demonstrated high recognition performance with the addition of BN and DO layers. CCNN and ICNN-BNDOA performance were compared. The statistical results showed that the proposed ICNN-BNDOA outperformed the CCNN with a training and testing accuracy of 0.6904 and 0.6861 respectively. It also outperformed with training and testing loss of 0.8910 and 0.9136 respectively.

Sun X., Li G., Qu P., Xie X., Pan X., Zhang W.

Traditional digital image processing methods extract disease features manually, which have low efficiency and low recognition accuracy. To solve this problem, In this paper, we propose a convolutional neural network architecture FL-EfficientNet (Focal loss EfficientNet), which is used for multi-category identification of plant disease images. Firstly, through the Neural Architecture Search technology, the network width, network depth, and image resolution are adaptively adjusted according to a group of composite coefficients, to improve the balance of network dimension and model stability; Secondly, the valuable features in the disease image are extracted by introducing the moving flip bottleneck convolution and attention mechanism; Finally, the Focal loss function is used to replace the traditional Cross-Entropy loss function, to improve the ability of the network model to focus on the samples that are not easy to identify. The experiment uses the public data set new plant diseases dataset (NPDD) and compares it with ResNet50, DenseNet169, and EfficientNet. The experimental results show that the accuracy of FL-EfficientNet in identifying 10 diseases of 5 kinds of crops is 99.72%, which is better than the above comparison network. At the same time, FL-EfficientNet has the fastest convergence speed, and the training time of 15 epochs is 4.7 h.

Khan A.I., Quadri S.M., Banday S., Latief Shah J.

• A suitable size expert-annotated apple leaf disease dataset has been prepared. • Presented a two-stage apple disease detection system based on Xception and Faster-RCNN. • Used a transfer learning method to initialize model by weight parameters learned on large-scale datasets. • Achieved an overall 88% of classification accuracy and our best detection model achieved mAP of 42%. • Promising results indicate that this system can be very helpful for farmers and Apple growers. Diseases and pests are one of the major reasons for low productivity of apples which in turn results in huge economic loss to the apple industry every year. Early detection of apple diseases can help in controlling the spread of infections and ensure better productivity. However, early diagnosis and identification of diseases is challenging due to many factors like, presence of multiple symptoms on same leaf, non-homogeneous background, differences in leaf colour due to age of infected cells, varying disease spot sizes etc. In this study, we first constructed an expert-annotated apple disease dataset of suitable size consisting around 9000 high quality RGB images covering all the main foliar diseases and symptoms. Next, we propose a deep learning based apple disease detection system which can efficiently and accurately identify the symptoms. The proposed system works in two stages, first stage is a tailor-made light weight classification model which classifies the input images into diseased, healthy or damaged categories and the second stage (detection stage) processing starts only if any disease is detected in first stage. Detection stage performs the actual detection and localization of each symptom from diseased leaf images. The proposed approach obtained encouraging results, reaching around 88% of classification accuracy and our best detection model achieved mAP of 42%. The preliminary results of this study look promising even on small or tiny spots. The qualitative results validate that the proposed system is effective in detecting various types of apple diseases and can be used as a practical tool by farmers and apple growers to aid them in diagnosis, quantification and follow-up of infections. Furthermore, in future, the work can be extended to other fruits and vegetables as well.

Thakur P.S., Sheorey T., Ojha A.

Crop diseases cause a substantial loss in the quantum and quality of agricultural production. Regular monitoring may help in early stage disease detection an d thereby reduction in crop loss. An automatic plant disease identification system based on visual symptoms can provide a smart agriculture solution to such problems. Various solutions for plant disease identification have been provided by researchers using image processing, machine learning and deep learning techniques. In this paper a lightweight Convolutional Neural Network ‘VGG-ICNN’ is introduced for the identification of crop diseases using plant-leaf images. VGG-ICNN consists of around 6 million parameters that are substantially fewer than most of the available high performing deep learning models. The performance of the model is evaluated on five different public datasets covering a large number of crop varieties. These include multiple crop species datasets: PlantVillage and Embrapa with 38 and 93 categories, respectively, and single crop datasets: Apple, Maize, and Rice, each with four, four, and five categories, respectively. Experimental results demonstrate that the method outperforms some of the recent deep learning approaches on crop disease identification, with 99.16% accuracy on the PlantVillage dataset. The model is also shown to perform consistently well on all the five datasets, as compared with some recent lightweight CNN models.

Ashwinkumar S., Rajagopal S., Manimaran V., Jegajothi B.

Agriculture is the major occupation in India and it loses 35% of the crop productivity annually owing to plant diseases. Earlier plant disease detection is a tedious process because of improper laboratory facilities and expert knowledge. Automated plant disease detection techniques are advantageous for reducing the laborious task of monitoring large crop farms and for identifying disease symptoms early on, i.e., when they appear on plant leaves. Recent advances in computer vision and deep learning (DL) models have demonstrated the value of developing automatic plant disease detection models based on visible symptoms on leaves. With this in mind, this article proposes an automated model for detecting and classifying plant leaf diseases using an optimal mobile network-based convolutional neural network (OMNCNN). The proposed OMNCNN model operates on different stages namely preprocessing, segmentation, feature extraction, and classification. It involves bilateral filtering (BF) based preprocessing and Kapur’s thresholding based image segmentation to identify the affected portions of the leaf image. In addition, the MobileNet model is applied as a feature extraction technique in which the hyperparameters are optimized by the use of emperor penguin optimizer (EPO) algorithm to enhance the plant disease detection rate. Finally, extreme learning machine (ELM) based classifier is utilized to allocate proper class labels to the applied plant leaf images. An extensive set of simulations were performed to highlight the superior performance of the OMNCNN model. The experimental outcome has shown promising results of the OMNCNN model over the recent state-of-art methods with the maximum precision of 0.985, recall of 0.9892, accuracy of 0.987, F-score of 0.985, and kappa of 0. 985.

Tariqul Islam M., Tusher A.N.

Day-by-day the cultivation of plants and albumen are increased speedily in order to fulfill the demand of human being and all the animals in this universe. Recently, the production rate of crop is abated due to different crop diseases. Agricultural scientists tried hard to finding the medication for the plant disorder. But the manual identification takes huge amount of time and less efficient. For the quick detection of plant disease different types of new technologies involvement with the cultivation sector bring as blessing. In this research work, deep learning process is used to diagnose the affliction and finding its cure through the images of transited leaf of “grape’’ and “strawberry”. In modern world, researchers can develop more accurate and efficient system for object detection and recognition using deep learning-based process. Here, we used convolutional-neural-network (CNN) algorithm to train the dataset where the accuracy rate is 93.63%. The farmer all over the world especially in Bangladesh can get the facilities form this work to the increment of production rate of grape and strawberry fruits through the reduction of disease and attack of insects.

Pandey A.K., Sinniah G.D., Babu A., Tanti A.

Tea (Camellia sinensis [L.] O. Kuntze) is a plantation crop, grown commercially in Asia, Africa, and South America. Among biotic threats to tea production, diseases caused by fungal pathogens are most significant. Worldwide, tea plants are challenged by several root, stem, and foliar diseases. Foliar diseases, blister blight, gray blight, and brown blight are particularly important as they adversely affect the bud and the two youngest leaves, causing loss of harvestable shoots. Over the past several decades, climate change and field management practices have influenced the risk of crop damage by several fungal pathogens, as well as the development and spread of diseases. Management interventions, such as the adoption of good cultural/agronomic practices, use of fungicides and microbial biocontrol agents, plant defense elicitors, and deployment of resistant cultivars, have mitigated damage to tea plants caused by fungal diseases. A clearer understanding of knowledge gaps and the benefits of plant disease management strategies available is needed. The present article reviews the prevailing knowledge of major fungal pathogens of the tea crop, their genetic variability, the damage they cause and its economic impact, and the need for new disease management strategies as climate change intensifies. We will also emphasize important knowledge gaps that are priority targets for future research.

Paymode A.S., Magar S.P., Malode V.B.

Mostly development of country depends of growth of agriculture sectors. Now a day agriculture is facing lot of challenges like unavailability of labor, drastic climate change, uncertainty in rain, natural disaster, different diseases on plant leaf and crops, no fixed prices and unavailability of markets and many more. But as the world continuously increasing the demands of food and more production needed in next 50 years. There are huge numbers of threats in agriculture field. The use of artificial intelligence technology, found best for the all agriculture challenges. So our proposed research focus on detecting and classifying the accurate type of diseases occurred on leaf at early stage. Our research aims to address the problem using the Deep Learning (DL) techniques. The AgroDeep mobile application developed for collection of real database of agriculture leafs and crop. The real diseased leaf images collected and captured through our mobile application. The captured images uploaded over database. There are total six types of crops leaf images (tomato, grapes, soybean, sugarcane, cotton and onion) collected. The tomato diseased leaf selected for detection and classification. The techniques supported whether diseases affected on leaf or not with percentage of accuracy. The best Convolution Neural Network (CNN) algorithm suitable for these analysis. The CNN based model gave the highest accuracy of 97 % which is highest forever for real captured diseased images. Our research playing exquisite role in agriculture sector and farmers. The proposed research supported to increase food production in the agriculture. Ultimately it gives more profit in the farming sector which motivate the farmers for agriculture.

Hu G., Wang H., Zhang Y., Wan M.

At present, the detection and severity estimation of tea diseases mainly rely on manual methods, which are time consuming and laborious. Existing machine learning and image processing methods used in disease detection and severity analysis of tea leaf blight (TLB) images captured in natural scenes have low accuracy because of the influence of light variation, shadow, varying shapes, and mutual occlusion of leaves. The current study proposes a deep learning method to improve the performance of detection and severity analysis of TLB. A Retinex algorithm is utilized to enhance the original images and reduce the influence of light variation and shadow. The TLB leaves are detected using a deep learning framework called Faster Region-based Convolutional Neural Networks, to improve the detection performance of blurred, occluded, and small pieces of diseased leaves. The detected TLB leaves are inputted into the trained VGG16 networks to achieve severity grading and facilitate disease severity analysis. Experimental results show that the detection average precision and the severity grading accuracy of the proposed method are improved by more than 6% and 9%, respectively, compared with the classical machine learning methods.

Gonzalez-Huitron V., León-Borges J.A., Rodriguez-Mata A.E., Amabilis-Sosa L.E., Ramírez-Pereda B., Rodriguez H.

• Deep transfer learning for disease detection in tomato leaves. • Evaluation and analysis from CNN models to select more suitable for a specific task. • Raspberry Pi 4 implementation for real-field operations. • GUI designed for easy usage. Deep learning has made essential contributions to classification and detection tasks applied to precision agriculture; however, it is vitally important to move towards an adoption of these techniques and algorithms through low-cost and low-consumption devices for daily use in crop fields. In this paper, we present the training and evaluation of four recent Convolutional Neural Networks models for the classification of diseases in tomato leaves. A subset of the Plantvillage dataset consisting of 18,160 RGB images has been divided into ten classes for transfer learning. The selected models have depthwise separable convolution architecture for application in low-power devices. Evaluation and analysis quantitatively and qualitatively is performed via quality metrics and saliency maps. Finally, an implementation on the Raspberry Pi 4 microcomputer with a graphical user interface is developed.

Kibriya H., Rafique R., Ahmad W., Adnan S.M.

The quality and quantity of the crop are significantly affected by numerous diseases in plants. In this regard, an early detection of such diseases is highly effective. Tomato is one of the important crops that is produced in large quantities with high commercial value. Several types of tomato diseases affect the crop at an alarming rate. In this paper, we deployed two Convolution Neural Network (CNN) based models i.e. GoogLeNet and VGG16 for tomato leaf disease classification. The proposed work aims to find the best solution to the problem of tomato leaf disease detection using a deep learning approach. VGG16 obtained 98% accuracy while GoogLeNet obtained 99.23% on Plant Village dataset containing 10735 leaf images. The proposed system can be used in tomato fields for early detection of disease to avoid production loss.

Sen S., Rai M., Das D., Chandra S., Acharya K.

Tea is one of the most consumed beverages and is produced from the tender leaves of the tea plant. Various biotic and abiotic factors are directly related to tea productivity. Among the biotic factors the most destructive one is the blister blight disease of tea caused by an obligate parasitic fungus Exobasidium vexans Massee. The pathogen attacks the tender leaves of the tea plant which directly interferes with the economic growth of the tea growing countries as tea has tremendous export value. Numerous studies have identified the symptoms, epidemiology of the pathogen and its control strategies.. Application of protectant and eradicant fungicides have shown promising results for controlling blister blight but overuse of chemical pesticides causes phytotoxicity, residual effects, thus use of microbial biocontrol agents are gaining more impetus. Different integrated disease management strategies along with modern emerging management approaches like elicitor mediated defense responses, development of transgenic tea plant, transcriptome study that induce many R-genes which ultimately provide innate immunity in tea plants. This review presents up-to-date information on blister blight disease which would help the future researchers to understand the host-pathogen interaction and the effective control measures to be adopt in a better way.

Agarwal M., Singh A., Arjaria S., Sinha A., Gupta S.

Tomato is the most popular crop in the world and in every kitchen, it is found in different forms irrespective of the cuisine. After potato and sweet potato, it is the crop which is cultivated worldwide. India ranked 2 in the production of tomato. However, the quality and quantity of tomato crop goes down due to the various kinds of diseases. So, to detect the disease a deep learning-based approach is discussed in the article. For the disease detection and classification, a Convolution Neural Network based approach is applied. In this model, there are 3 convolution and 3 max pooling layers followed by 2 fully connected layer. The experimental results shows the efficacy of the proposed model over pre-trained model i.e. VGG16, InceptionV3 and MobileNet. The classification accuracy varies from 76% to 100% with respect to classes and average accuracy of the proposed model is 91.2% for the 9 disease and 1 healthy class.

Ferdouse Ahmed Foysal M., Shakirul Islam M., Abujar S., Akhter Hossain S.

Neural networks have achieved a significant result in every arena of information technology, where it has been used to solve any major problem. As a consequence, upward inclination of development has been observed in those industries. Nowadays, because of using NN, the challenges became easier than before to deal with. The agriculture industry is one of the important and largest industries, where technology can make a major contribution by solving their certain problems. Implementation of artificial intelligence can make this industry more successful and faster growing. Since the very beginning, plant diseases are one of the major factors behind low-quality products. So, through identifying those diseases earlier, we can make a great contribution to this agroindustry. Therefore, in this work, a definite detection of tomato diseases has been presented. Several existing and proposed method of identifying the disease through analyzing tomato leaves has been discussed. We have proposed a 15-layered Deep Convolutional Neural Network. Basically, this research will state a basic approach for tomato disease classification. This will be able to classify five different tomato diseases, and the proposed model has achieved fairly high accuracy with low cross-entropy rate. Several simulation results have been measured and discussed.

Panchbhai K.G., Lanjewar M.G.

The identification of tea leaf diseases holds considerable significance for preserving the health of tea plants and preventing losses in tea production. This study introduced a hybrid framework by combining modified state-of-the-art (SOTA) models with feature selection and Machine Learning (ML) classifiers for recognizing four types of tea leaf diseases. The investigation utilized SOTA models, namely VGG16, Xception, and ResNet152V2. These architectures underwent modification by adding extra layers, serving as feature extractors from tea leaves. The extracted features then underwent a feature selection process to identify the most relevant ones, which were subsequently employed in ML classifiers for predicting tea leaf diseases. The proposed method demonstrated outstanding performance with a 2-fold average accuracy of 99.5%, an Area Under the Curve (AUC) of 1.0, and a p value of 0.001.

Vermelho A.B., Moreira J.V., Akamine I.T., Cardoso V.S., Mansoldo F.R.

Pesticide use in crops is a severe problem in some countries. Each country has its legislation for use, but they differ in the degree of tolerance for these broadly toxic products. Several synthetic pesticides can cause air, soil, and water pollution, contaminating the human food chain and other living beings. In addition, some of them can accumulate in the environment for an indeterminate amount of time. The agriculture sector must guarantee healthy food with sustainable production using environmentally friendly methods. In this context, biological biopesticides from microbes and plants are a growing green solution for this segment. Several pests attack crops worldwide, including weeds, insects, nematodes, and microorganisms such as fungi, bacteria, and viruses, causing diseases and economic losses. The use of bioproducts from microorganisms, such as microbial biopesticides (MBPs) or microorganisms alone, is a practice and is growing due to the intense research in the world. Mainly, bacteria, fungi, and baculoviruses have been used as sources of biomolecules and secondary metabolites for biopesticide use. Different methods, such as direct soil application, spraying techniques with microorganisms, endotherapy, and seed treatment, are used. Adjuvants like surfactants, protective agents, and carriers improve the system in different formulations. In addition, microorganisms are a tool for the bioremediation of pesticides in the environment. This review summarizes these topics, focusing on the biopesticides of microbial origin.

Khan B., Das S., Fahim N.S., Banerjee S., Khan S., Al-Sadoon M.K., Al-Otaibi H.S., Islam A.R.

Manual identification of tomato leaf diseases is a time-consuming and laborious process that may lead to inaccurate results without professional assistance. Therefore, an automated, early, and precise leaf disease recognition system is essential for farmers to ensure the quality and quantity of tomato production by providing timely interventions to mitigate disease spread. In this study, we have proposed seven robust Bayesian optimized deep hybrid learning models leveraging the synergy between deep learning and machine learning for the automated classification of ten types of tomato leaves (nine diseased and one healthy). We customized the popular Convolutional Neural Network (CNN) algorithm for automatic feature extraction due to its ability to capture spatial hierarchies of features directly from raw data and classical machine learning techniques [Random Forest (RF), XGBoost, GaussianNB (GNB), Support Vector Machines (SVM), Multinomial Logistic Regression (MLR), K-Nearest Neighbor (KNN)], and stacking for classifications. Additionally, the study incorported a Boruta feature filtering layer to capture the statistically significant features. The standard, research-oriented PlantVillage dataset was used for the performance testing, which facilitates benchmarking against prior research and enables meaningful comparisons of classification performance across different approaches. We utilized a variety of statistical classification metrics to demonstrate the robustness of our models. Using the CNN-Stacking model, this study achieved the highest classification performance among the seven hybrid models. On an unseen dataset, this model achieved average precision, recall, f1-score, mcc, and accuracy values of 98.527%, 98.533%, 98.527%, 98.525%, and 98.268%, respectively. Our study requires only 0.174 s of testing time to correctly identify noisy, blurry, and transformed images. This indicates our approach's time efficiency and generalizability in images captured under challenging lighting conditions and with complex backgrounds. Based on the comparative analysis, our approach is superior and computationally inexpensive compared to the existing studies. This work will aid in developing a smartphone app to offer farmers a real-time disease diagnosis tool and management strategies.