Crop development and yield are crucial factors that affect agriculture and farmers in every conceivable way, including economically, socially, and politically (Production et al., 2014). Consequently, monitoring the development of crops to detect various types of illness is a crucial step at specific times. However, naked human eye may not be sufficient, and occasionally deceptive scenarios may occur (Arsenovic et al., 2019). Automatic recognition and classification of diverse agricultural diseases are required for accurate identification. This paper provides an overview of the methods proposed for our research project. This paper contains the suggested methodology’s context, problem definition, objectives, and scope. Pakistani farmers’ illiteracy is one of the major contributors to a rise in microbial infections (Government of Pakistan, 2021). Once a disease has infected a crop, it is difficult for farmers to determine its root cause. Pathogens and pests are wreaking havoc on crops. This information comes from a study report published by UC Agriculture and Natural Resources; the crop increases the yield of five major food crops by 10 to 40% (Ali et al., 2017). In the context of Pakistan, where agriculture provides 16% of the GDP and employs over 60% of the people, it is essential to adopt extensive steps to prevent plant diseases. According to the Ministry of Food Processing Industries, agricultural losses in 2016 totaled thirteen billion US dollars (GoI, 2019). Image processing and neural networks can be used to perform one of the beneficial steps in plant disease diagnosis techniques (Tugrul et al., 2022). Recent research has demonstrated that neural networks and deep learning perform categorization tasks effectively.

Agriculture is a crucial sector in nations such as Pakistan, whose economies depend directly or indirectly on agriculture. It highlights the requirement of caring for plants from the seedling stage to the harvest. To produce the desired yield, the crop plants must endure sometimes unfavourable meteorological conditions, survive various diseases, and attacks by animals. The latter can be resolved if the crops are shielded from the various animals. Weather circumstances are beyond human control but there are technologies to mitigate abiotic stresses (Kerchev et al., 2020; Talaviya et al., 2020). Lastly, it is vital to safeguard the crop against various illnesses, as they can affect the crop’s total growth and output. If these diseases can be identified promptly, the crop can be safeguarded with the necessary agrochemicals. Digitalizing this disease diagnosis and classification procedure might be advantageous for farmers. It will reduce the time and accuracy required to identify and classify diseases.

Agriculture is the most ancient occupation that occurred even before science. With the advancement of science, it became clear that plants were living organisms capable of respiration, reproduction, and susceptible to certain diseases. These diseases, caused by many microorganisms, including bacteria, viruses, and fungi, diseases can cause significant damage to crops and can harm humans, as it is the case with pathogen-produced mycotoxins and other toxicants (Nan et al., 2022). Furthermore, destruction of crops by pathogens can cause human starvation. In 1840, a disease known as the Late blight destroyed a substantial potato crop. It is also known as the Irish famine, and it was a sad period in European history during which many people perished from hunger (Goss et al., 2014). As plants are vital to human survival, so we must safeguard them against deforestation and numerous plant diseases. A significant portion of Pakistan’s population is still engaged in agriculture. It ranks second in agricultural production worldwide. However, two major issues harm Pakistan’s crops: natural disasters and infectious diseases. According to United Nations data, agriculture lost 96 billion dollars in 2005-2015. We have no control over natural calamities, but we can control plant illnesses caused by microorganisms. In the context of Pakistan, where agriculture provides 16% of the GDP and employs over 60% of the people, it is essential to adopt extensive steps to prevent plant diseases. According to the Ministry of Food Processing Industries, agricultural losses in 2016 totaled thirteen billion dollars. Image processing and neural networks can be used to perform one of the beneficial steps in plant disease diagnosis techniques. Recent research has demonstrated that neural networks and deep learning perform categorization tasks effectively.

The proposed research work focuses on three plant species: potato, tomato, and pepper. In a poll, 61.33 percent of potato producers cited light as one of the primary causes of crop loss. In 2020, according to national statistics, almost 60 percent of tomato crops in Punjab failed to owe to a virus. Pakistan accounts for 40 percent of the world’s total pepper production. Additionally, pepper has numerous natural benefits for throat infections. China is the largest producer of tomatoes and potatoes, whereas India is the second-highest producer of these two crops. Enhancing fertilization and automating the disease detection system can increase agricultural yield in our country.

To our knowledge, plant leaf lesion segmentation, lesion subtype categorization, and overall survival prediction have been addressed individually without regard for their inherent linkages. This research uses deep neural networks along with some advanced machine learning techniques to provide a complete strategy for plant leaf lesion segmentation, the recognition of lesion subtypes, and the survival estimation of a plant. Detailed contributions are listed below. First, a novel context-aware Convolutional Neural Network (CNN)-based technique for plant lesion segmentation is presented. In the second phase, a hybrid model is utilized for plant survival estimation using the segmented ROI information. A context-encoded convolutional neural network (CANet) (Zhang et al., 2019) is employed to extract high-dimensional features which are classified by the linear regression machine learning methods to make plant life survival predictions. In the final phase of the proposed framework, all the distinct tasks, i.e., segmentation, classification, and survival estimation, are merged into a single interconnected deep learning strategy. In conclusion, whereas the plant damage tests and classification criteria recommend foliar and pathological images, the suggested method effectively detects plant disease using only leaf RGB image data. On the PlantVillage dataset, the proposed segmentation and classification scheme was validated.

The rest of this paper is organized as follows. Section 2 shows the literature review of Plant Lesion Segmentation using deep learning methods. Section 3 presents the methodology of the proposed models. Section 4 shows the experimental results. Finally, Section 5 concluded the proposed work.

Using a mix of a Deep Learning classification model (CNN) and a features selection method genetic algorithm (GA), a model is presented for the diagnosis and recognition of tomato plant disease using the leaf image data (Tugrul et al., 2022). The proposed given framework was trained on 500 images belonging to 4 types of diseases. The features learning block of the CNN model is used to extract important visual characteristics and for classification. In a research study, an examination of the efficacy of CNN architecture for the recognition of plant diseases using the leaf data was conducted to detect diseases in soybean plants (Yu et al., 2022), the framework is implemented using LeNet which is the smallest and simplest CNN architectures. The leaf photos of 13,842 images of soybean plants are gathered from the PlantVillage benchmark dataset. The above framework has an accuracy of 98.44%, demonstrating CNN’s usefulness for classifying plant diseases based on leaf images. The approach to plant disease identification involves the construction of a modern model for identifying 13 plant illnesses from photographs of healthy plant leaves (Sladojevic et al., 2016). Caffe, an architecture for deep learning, was used to train the data. The framework in question yielded outcomes with a 91 to 98 percent accuracy. The author of the research article (He et al., 2020) developed a two-stage approach. In the beginning, the architectures based on meta information of Regional Convolutional Neural Network (R-CNN), Regional Fully Connected Network (R-FCN), and Single Shot detector (SSD) are combined to develop a single object detector. The features learning blocks of VGG16 (Khattab, 2019), VGG-19 (Szymak et al., 2020), and inception-V3 (Szymak et al., 2020) are utilized to extract high-dimension features from the training data and evaluation of models performances. Comparing the proposed model to other similar detection models, the proposed model is found more time efficient. A novel deep-learning architecture is developed for the detection and recognition of mango plant diseases (Saleem et al., 2021). The proposed CNN is trained on 754 unhealthy and 780 healthy mango leaf image data. The custom framework achieves an average f1-score of 97.01%. The author (Durmus et al., 2017) suggested a system based on the architecture of convolutional neural networks to recognize and categorize several potato plant diseases. The dataset utilized for this framework contains 2,465 photos of potatoes. The author (Geetharamani and J., 2019) researched and recognized the benefits and cons of the model and the performance of deep learning neural networks, which are used to recognize and classify various plant diseases. The literature review and picture database experiments comprise 50,000 photographs of many plant diseases (Durmus et al., 2017). The author presented a novel deep learning-based framework that is capable of segmenting the affected region on the leaf and recognizing the type of disease in tomato plants (Islam et al., 2022). This framework’s dataset comprises 13,281 tomato leaf photos with nine types of illnesses which are collected from the PlantVillage dataset. The model achieved an average of 99.91% accuracy on the training data while on the testing data the average accuracy achieved is 98.96%. The author (Goss et al., 2014) concentrated on identifying and categorizing diverse diseases in rice plants using the CNN features and SVM classifier for decision-making. of a framework utilizing CNN architecture and SVM. The training data consists of 696 rice plant leaf images which belong to four types of rice plant diseases. Maximum accuracy of 91.37 percent is reached when evaluating the accuracies of diverse training and testing datasets. In the case of sugar beet, an existing model is upgraded, utilizing the faster region-based CNN architecture by modifying the parameters for recognizing disease-affected regions (Nasirahmadi et al., 2021). The dataset comprises 155 photos of sugar beets, and the proposed framework attained an accuracy rate of 95.48 percent. In the case of Olive plant diseases, the author of the research article (He et al., 2020) compared a transfer learning scenario with CNN architectures such as VGG-16 and VGG-19, as well as proposed CNN architectures (Alshammari et al., 2022). The framework applied to the dataset includes around 3,400 photos of Olive plant leaves. In this framework, a data augmentation technique was utilized to increase the size of the dataset. Before data augmentation, the accuracy was approximately 88%; after data augmentation, it was approximately 95%. The study paper (Abdulridha et al., 2020) proposes a CNN-based tailored model for tomato leaf disease detection. In addition, I compared the proposed model to models of machine learning and VGG-16. The proposed model achieved 98.4% accuracy, the KNN model achieved 94.9% accuracy, and the VGG-16 model achieved 93.5% accuracy. The dataset of tomato leaf pictures utilized by this framework is derived from the Plant village dataset. Deep learning’s transfer learning technique is used to detect and categorize illnesses using leaf images of two crops, such as cucumber and rice (Liu and Wang, 2021). The suggested framework was applied to 2,430 photos of cucumbers and rice afflicted with eight illnesses collected from the plant village dataset.

The proposed framework achieved a precision of 90.84 percent (Liu and Wang, 2021). The influence of deep learning on diagnosing plant illnesses using leaf photos was examined. CNN architecture functions as a black box model for plant disease diagnosis. Also covered are the many hyperparameter characteristics that affect classification accuracy. Numerous models and research have used deep learning scenarios to identify and classify illnesses in certain plant categories. Deep learning can also be used to identify and categorize the macronutrients present in a particular plant. The proposed technique for monitoring plant health checks several stages from the seedling stage through the yielding stage to increase yield. The suggested system was implemented using a dataset of 571 photos, including images of tomato leaves and tomato fruit at various stages of crop growth. The inception-ResNet v2 and autoencoder performance was 87.27 percent and 79.09 percent, respectively. This literature demonstrates the impact of transfer learning on identifying and categorizing plant diseases using photographs of leaf surfaces. According to the article’s author (Siddiqua et al., 2022), picture segmentation with the aid of colors, i.e., color image segmentation approaches, improves comprehension and problem-solving. One can first determine an image’s red, Green, and Blue color components. The red and green components aid in identifying the yellow portions of the image, typically indicated as infected. Fuzzy logic is an effective method for solving disease classification issues (Sibiya and Sumbwanyambe, 2021), the author proposes the minimum distance approach, a genetic algorithm modification, to locate a plant’s infected portion for picture segmentation (Ngugi et al., 2021). After picture segmentation, the author examined the accuracy of the technique using different classification algorithms, such as k mean clustering and SVM (Bargelloni et al., 2021). In this paper, the author uses a convolution neural network technique to diagnose various plant illnesses (Production et al., 2014; Saleem et al., 2021). The author has conducted an exhaustive study. Photographs of diverse plant leaves, including images of both sick and healthy leaves, are captured. The author has grouped it into numerous categories, and all CNN designs achieved an accuracy greater than 97%. AlexNet (Yoo et al., 2021), AlexNetOWTBn, GoogLeNet (Wang et al., 2015), Overfeat, and VGG are the CNN architectures. The author has comprehensively analyzed several deep learning algorithms, their benefits and drawbacks, and optimization strategies (Sarker, 2021). In the linked work, these strategies have also been compared. In this publication (Ngugi et al., 2021), the author describes the proposed algorithm in depth, the image acquisition was the initial step, followed by image enhancement and segmentation. The HSV approach was utilized for the segmentation of color images. Integrated into the instrument for evaluating plant disease were sensors that could determine the meteorological and climatic factors affecting the plant disease in real-time.

This section includes a comprehensive overview of the proposed DCNN model’s architecture and training method, including the preparation of the dataset and experimental procedures. The suggested model for detecting plant leaf diseases begins with dataset preparation and concludes with model prediction. Python 3.8, TensorFlow Library version 2.10.0, NumPy 1.23.4, matplotlib 3.6.1, and OpenCV 4.6.0 are used to prepare the training dataset and implement the proposed DCNN model, respectively. The simulations, i.e., model development, training, validation, etc., are performed on an HP Z440 workstation consisting of core i7 12 cores of CPU and a DDR4 ram of 48 GB. The proposed scheme also utilized NVidia RTX-3090 Graphical Processor Unit (GPU), which uses the CUDA framework to allow the parallel processing speeds up the proposed model training and testing procedure. The workstation for implementing the proposed DCNN is equipped with a dual Intel Xeon Silver 4310 (12 cores, 24 threads, and 2.10Ghz) processor and six Nvidia Tesla P100 GPUs to expedite the training of deep neural networks. The following sections will explain all the important phases of the proposed plant disease detection framework in detail. The section that follows addresses the specifics of data set preparation and preprocessing.

This section gives a comprehensive summary of the experiments’ outcomes to determine the proposed technique’s evaluation capabilities. This section also describes the dataset used to evaluate performance. The proposed framework is implemented in Python and runs on systems with Nvidia RTX 3090. CenterNet configuration on the PlantVillage dataset for classifying and scoring plant leaf diseases.

Deep learning-based algorithms have attained state-of-the-art performance in numerous sectors where they have been widely implemented. However, leaf lesion segmentation provides numerous specific challenges:

The fundamental challenge with disease classification is the lack of data. Even after increasing the training sample size using data augmentation approaches, 110250 examples may not be sufficient for deep learning in this work. Similar data deficiency issues exist for global survival prediction. In the PlantVillage Challenge training phase, only ten classes are accessible. In addition to the deep learning-based approach, we implement global survival prediction using a conventional machine learning method by extracting features such as gray-level co-occurrence matrix (GLCM), intensity, etc., applying LASSO to select features and then using linear regression for survival prediction. We compare the outcome to our proposed method’s outcome. The comparison demonstrates that the performance of the proposed strategy is superior. In this paper, we also assess the influence of various diseases on overall survival. There are three classification models trained for the classes of peppers, potatoes, and tomatoes. The detection accuracy of models for pepper, potato, and tomato plants is 99.11%, 94.01%, and 99.04%, respectively shown in Table 12 . The proposed deep-learning model for pepper, potato, and tomato plant disease detection is shown in Figure 10 .

This study investigates several plant disease diagnostic and analysis tasks using deep learning and plant leaf imagery. These tasks involve segmenting the leaf lesion area, classifying the lesion into its subtypes, and predicting the plant’s overall survival. We built a context-aware 3D CNN that extracts and classifies high-dimensional, non-invariant characteristics from a plant leaf image to identify the disease type. Similarly, a unique method is established utilizing the regression model to predict long-term, short-term, and intermediate-term plant survival. The features learning block of the CANet CNN model extracts features, reduces the dimension of the features vector by picking only the optimum features, and employs the LASSO features selection algorithm. The PlantVillage Dataset comprises numerous photos of crop leaf diseases. This study has validated the suggested model using three different plant diseases: pepper, potato, and tomato. The pepper plant has only two classes, but the potato and tomato plants have multiple classes. The suggested model achieves a DICE coefficient of 90% while segmenting plant leaf lesions. The classification accuracy for detecting pepper illness is 91.11%, for detecting potato disease is 93.01%, and for detecting tomato, the disease is 99.04%. Consequently, the improved accuracy suggests that the suggested method applies to the PlantVillage dataset and other datasets for lesion segmentation, classification, and plant survival calculation.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/supplementary material.

MS, BS, TH, AU, AA, IS, TG, and FA conceptualized this study, conducted experiments, wrote the original draft, revised the manuscript, performed the data analysis, supervised the work, and revised the manuscript. FarA designed the experimental plan, supervised the work, and revised the manuscript. All authors have read and agreed to the published version of the manuscript.