Figure - available from: Frontiers in Plant Science
This content is subject to copyright.
Example of leaf images from the PlantVillage dataset, representing every crop-disease pair used. (1) Apple Scab, Venturia inaequalis (2) Apple Black Rot, Botryosphaeria obtusa (3) Apple Cedar Rust, Gymnosporangium juniperi-virginianae (4) Apple healthy (5) Blueberry healthy (6) Cherry healthy (7) Cherry Powdery Mildew, Podoshaera clandestine (8) Corn Gray Leaf Spot, Cercospora zeae-maydis (9) Corn Common Rust, Puccinia sorghi (10) Corn healthy (11) Corn Northern Leaf Blight, Exserohilum turcicum (12) Grape Black Rot, Guignardia bidwellii, (13) Grape Black Measles (Esca), Phaeomoniella aleophilum, Phaeomoniella chlamydospora (14) Grape Healthy (15) Grape Leaf Blight, Pseudocercospora vitis (16) Orange Huanglongbing (Citrus Greening), Candidatus Liberibacter spp. (17) Peach Bacterial Spot, Xanthomonas campestris (18) Peach healthy (19) Bell Pepper Bacterial Spot, Xanthomonas campestris (20) Bell Pepper healthy (21) Potato Early Blight, Alternaria solani (22) Potato healthy (23) Potato Late Blight, Phytophthora infestans (24) Raspberry healthy (25) Soybean healthy (26) Squash Powdery Mildew, Erysiphe cichoracearum (27) Strawberry Healthy (28) Strawberry Leaf Scorch, Diplocarpon earlianum (29) Tomato Bacterial Spot, Xanthomonas campestris pv. vesicatoria (30) Tomato Early Blight, Alternaria solani (31) Tomato Late Blight, Phytophthora infestans (32) Tomato Leaf Mold, Passalora fulva (33) Tomato Septoria Leaf Spot, Septoria lycopersici (34) Tomato Two Spotted Spider Mite, Tetranychus urticae (35) Tomato Target Spot, Corynespora cassiicola (36) Tomato Mosaic Virus (37) Tomato Yellow Leaf Curl Virus (38) Tomato healthy.
Source publication
Crop diseases are a major threat to food security, but their rapid identification remains difficult in many parts of the world due to the lack of the necessary infrastructure. The combination of increasing global smartphone penetration and recent advances in computer vision made possible by deep learning has paved the way for smartphone-assisted di...
Similar publications
Agriculture is backbone of India and major revenue producing sector. But change in climate and its unpredictability directly affect the production and maintenance of crops. Furthermore, condition becomes even worst when the crops are infected by any disease which leads to crop failure. Farmer Suicidal Rate is 11.2% of all suicides in India. Increas...
Detection and classification of leaf and crop diseases in a traditional way is a very laborious task as it involves a significant amount of physical work, huge expert manpower, and valuable time. Automatic systems are more accurate and require less time, labor, and physical work. Artificial intelligence and deep learning-based systems can help in t...
Introduction/Background
Plant diseases and pernicious insects are a considerable threat in the agriculture sector. Leaf diseases impact agricultural production. Therefore, early detection and diagnosis of these diseases are essential. This issue can be addressed if a farmer can detect the diseases properly.
Objective
The fundamental goal of this p...
The identification of crop diseases is one of the major concerns that the agricultural industry has to deal with. The detection and classification of leaves is essential in agriculture, forestry, rural medicine, and other commercial applications, among other things. The diagnosis of sugar cane plant leaf disease is required for automatic weed ident...
The research is devoted to the analysis of the current state of chemical crop protection market in Russia. Significant industry growth has occurred in last decade due to the attractiveness of agriculture as an export-oriented industry. Russian agricultural products are in stable demand all over the world. Countries in the Middle East, South-Eastern...
Citations
... These datasets still serve as the mainstay in many studies, but most of them are taken indoors. With further research, Mohanty et al. [83,84] pointed out that models based solely on laboratory data cannot withstand inspection of detection in wheat fields. However, with the peak of research on the automatic detection of wheat FHB, more pluralistic data have started their applications: multiple varieties [85], omnidirectional [86], multiple weather conditions and backgrounds [87], multiple raters and years [88], multiple growth stages and image resolutions [89], etc. ...
Fusarium head blight (FHB) is a major threat to global wheat production. Recent reviews of wheat FHB focused on pathology or comprehensive prevention and lacked a summary of advanced detection techniques. Unlike traditional detection and management methods, wheat FHB detection based on various imaging technologies has the obvious advantages of a high degree of automation and efficiency. With the rapid development of computer vision and deep learning technology, the number of related research has grown explosively in recent years. This review begins with an overview of wheat FHB epidemic mechanisms and changes in the characteristics of infected wheat. On this basis, the imaging scales are divided into microscopic, medium, submacroscopic, and macroscopic scales. Then, we outline the recent relevant articles, algorithms, and methodologies about wheat FHB from disease detection to qualitative analysis and summarize the potential difficulties in the practicalization of the corresponding technology. This paper could provide researchers with more targeted technical support and breakthrough directions. Additionally, this paper provides an overview of the ideal application mode of the FHB detection technologies based on multi-scale imaging and then examines the development trend of the all-scale detection system, which paved the way for the fusion of non-destructive detection technologies of wheat FHB based on multi-scale imaging.
... These AI-driven solutions empower farmers with essential tools and insights for informed decision-making and optimal resource use. Current research focuses on advanced computational techniques like data mining and neural networks, which are transforming agriculture through applications such as predictive analytics and disease detection [14]. Multi-Agent Systems (MAS) have been shown to enhance the accuracy and comprehensiveness of biomedical literature searches. ...
... Therefore, the agricultural protection landscape is on the verge of a new era. Embarking on a journey situated at the convergence of agriculture and artificial intelligence, the emerging sentinel in this domain is not a human but rather a sophisticated amalgamation of data, algorithms and sensors (Mohanty et al., 2016). The relentless march of technology has endowed humanity with the tools to transform soybean fields into digitally fortified fortresses, replete with a network of vigilant electronic eyes and ears (Ferentinos, 2018). ...
Background: Soybean cultivation faces challenges from pest infestations, necessitating advanced and proactive pest management strategies. Traditional methods often lag in early detection, resulting in substantial crop losses. This study addresses this gap by employing machine learning, specifically sequential convolutional neural networks (CNNs) and recurrent neural networks (RNNs), to transform early pest detection in soybean fields. The approach leverages a comprehensive dataset comprising high-resolution crop images and environmental variables, offering insights into soybean ecosystems. Methods: Using CNNs for picture processing and RNNs for collecting temporal relationships in environmental factors, a complex deep learning model is created. The dataset contains 1050 images of soybeans including pests such as Anticarsia, Coccinellidae and without pests (healthy). The model is trained for 20 epochs. The model has been carefully validated for accuracy, sensitivity and efficacy in early insect identification. The diversity of the dataset ensures that the model may be adjusted to a range of soybean growing situations. Result: The outcomes demonstrate the model’s unparalleled effectiveness, routinely outperforming conventional techniques with an accuracy rate of 95%. Its exceptional sensitivity reduces financial and environmental expenses, highlighting its versatility in a range of soybean growing environments. In light of the difficult global agricultural landscape, this study offers a novel strategy for proactive and sustainable pest management, which is essential to guaranteeing strong soybean crop yields.
... In a recent study by [5], the ResNet50 model was utilized to detect five strawberry diseases in Miaoli County, Taiwan. Meanwhile, Fuentes et al. [47] explored the effectiveness of DL metaarchitectures like Faster R-CNN, R-FCN, and SSD in detecting nine types of diseases and pests amidst complex surroundings. Further contributions in the realm of plant disease recognition come from studies by [44], among others, underscoring the growing significance of DL in addressing agricultural challenges. ...
... Meanwhile, [49] combined transfer learning with popular Convolutional Neural Network (CNN) architectures such as VGGNet, AlexNet, and GoogLeNet to recognize plant types, securing the third position in PlanCLEF2016. In a study by [47], an improved inception V2 architecture was leveraged to identify plant species. Experimental results in real-world scenarios demonstrated the method's superior accuracy compared to Faster RCNN in identifying leaf species amidst complex environments. ...
Agriculture is a vital part of the global economy, and recent breakthroughs in deep learning technology are showing great potential to revolutionize this industry. This paper takes a close look at the expanding research on how deep learning and machine learning are being used to tackle agricultural challenges, especially through image processing and computer vision. The paper explores various deep neural network architectures and machine learning methods currently being used in agriculture, offering a summary of the latest advancements. It highlights the wide range of applications for deep learning in this field, such as managing irrigation systems, detecting weeds, recognizing patterns, and identifying crop diseases. In addition, the paper delves into the specific deep learning models utilized, the data sources they rely on, the metrics used to measure their performance, and the hardware that supports these technologies. It also considers the potential for real time applications, particularly with autonomous robotic platforms. The survey underscores that deep learning techniques significantly outperform traditional machine learning methods in terms of accuracy, showcasing their superior ability to enhance agricultural practices.
... Our experiment setup for REP extends the methodology described in Sec. 3 by including Split PlantDisease [50] dataset for image classification. Split PlantDisease contains 54,303 plant disease images across 38 categories. ...
On-device continual learning (CL) requires the co-optimization of model accuracy and resource efficiency to be practical. This is extremely challenging because it must preserve accuracy while learning new tasks with continuously drifting data and maintain both high energy and memory efficiency to be deployable on real-world devices. Typically, a CL method leverages one of two types of backbone networks: CNN or ViT. It is commonly believed that CNN-based CL excels in resource efficiency, whereas ViT-based CL is superior in model performance, making each option attractive only for a single aspect. In this paper, we revisit this comparison while embracing powerful pre-trained ViT models of various sizes, including ViT-Ti (5.8M parameters). Our detailed analysis reveals that many practical options exist today for making ViT-based methods more suitable for on-device CL, even when accuracy, energy, and memory are all considered. To further expand this impact, we introduce REP, which improves resource efficiency specifically targeting prompt-based rehearsal-free methods. Our key focus is on avoiding catastrophic trade-offs with accuracy while trimming computational and memory costs throughout the training process. We achieve this by exploiting swift prompt selection that enhances input data using a carefully provisioned model, and by developing two novel algorithms-adaptive token merging (AToM) and adaptive layer dropping (ALD)-that optimize the prompt updating stage. In particular, AToM and ALD perform selective skipping across the data and model-layer dimensions without compromising task-specific features in vision transformer models. Extensive experiments on three image classification datasets validate REP's superior resource efficiency over current state-of-the-art methods.
... However, one of the primary challenges in recognizing stressed plants in agriculture is the scarcity of data. While several datasets exist for tasks like plant disease detection, such as PlantVillage [16], and plant identification [6], these datasets are often unsuitable for water-stressed plant classification. The reason lies in the absence of classes representing drought-stressed plants, and some datasets consist of leaf images that do not reflect real-world images as those captured by mobile robots. ...
The detection of water-stressed plants have a significant impact in the agricultural field providing water conservation and improved crop yield. Deep Learning (DL) models can play a crucial role in this task based on the CNN models for image classification. A challenge is to build a dataset containing water-stressed plant images and non water-stressed plant images in order to train the classification model while the absence of such an open source dataset. In this paper, we compare between a two augmentation techniques: The first is based on Latent Diffusion Models (LDMs), and the second is based on a novel architecture entitled NST-SAM exploiting Segment Anything Model (SAM) and Neural Style Transfer (NST). In our comparative approach, we have created 4 datasets: The first composed of real plant images collected from the internet, the second and third are an augmentation of the first using respectively LDMs and NST-SAM, and the last with 1250 images containing the real and all augmentation images (LDMs and NST-SAM). The classification metrics indicates that when associating the LDMs and NST-SAM we augment the performances of the model presenting an accuracy of 84% using VGG16 and 85% with ResNet50.
... The system got accuracy up to 96.7%. Furthermore, AlexNet and GoogLeNet architectures for deep learning were used by Mohanty et al. (Mohanty et al. 2016) to produce models for categorizing tomato leaf diseases. By combining learning methods and different training and testing splits, their system got an accuracy of 99.35% using the PlantVillage (PlantVillage Dataset 2023) dataset. ...
Farmers face the formidable challenge of meeting the increasing demands of a rapidly growing global population for agricultural products, while plant diseases continue to wreak havoc on food production. Despite substantial investments in disease management, agriculturists are increasingly turning to advanced technology for more efficient disease control. This paper addresses this critical issue through an exploration of a deep learning-based approach to disease detection. Utilizing an optimized Convolutional Neural Network (E-CNN) architecture, the study concentrates on the early detection of prevalent leaf diseases in Apple, Corn, and Potato crops under various conditions. The research conducts a thorough performance analysis, emphasizing the impact of hyperparameters on plant disease detection across these three distinct crops. Multiple machine learning and pre-trained deep learning models are considered, comparing their performance after fine-tuning their parameters. Additionally, the study investigates the influence of data augmentation on detection accuracy. The experimental results underscore the effectiveness of our fine-tuned enhanced CNN model, achieving an impressive 98.17% accuracy in fungal classes. This research aims to pave the way for more efficient plant disease management and, ultimately, to enhance agricultural productivity in the face of mounting global challenges. To improve accessibility for farmers, the developed model seamlessly integrates with a mobile application, offering immediate results upon image upload or capture. In case of a detected disease, the application provides detailed information on the disease, its causes, and available treatment options.
... The battle against cotton diseases endures, with ongoing efforts to avert crop losses by early and effective disease detection, followed by timely intervention (Mohanty et al., 2016;Guo et al., 2022). While manual disease detection is the prevailing approach, it is hampered by reliability issues and is impractical for large-scale monitoring due to time and cost constraints (Peyal et al., 2022). ...
Cotton, a vital textile raw material, is intricately linked to people’s livelihoods. Throughout the cotton cultivation process, various diseases threaten cotton crops, significantly impacting both cotton quality and yield. Deep learning has emerged as a crucial tool for detecting these diseases. However, deep learning models with high accuracy often come with redundant parameters, making them challenging to deploy on resource-constrained devices. Existing detection models struggle to strike the right balance between accuracy and speed, limiting their utility in this context. This study introduces the CDDLite-YOLO model, an innovation based on the YOLOv8 model, designed for detecting cotton diseases in natural field conditions. The C2f-Faster module replaces the Bottleneck structure in the C2f module within the backbone network, using partial convolution. The neck network adopts Slim-neck structure by replacing the C2f module with the GSConv and VoVGSCSP modules, based on GSConv. In the head, we introduce the MPDIoU loss function, addressing limitations in existing loss functions. Additionally, we designed the PCDetect detection head, integrating the PCD module and replacing some CBS modules with PCDetect. Our experimental results demonstrate the effectiveness of the CDDLite-YOLO model, achieving a remarkable mean average precision (mAP) of 90.6%. With a mere 1.8M parameters, 3.6G FLOPS, and a rapid detection speed of 222.22 FPS, it outperforms other models, showcasing its superiority. It successfully strikes a harmonious balance between detection speed, accuracy, and model size, positioning it as a promising candidate for deployment on an embedded GPU chip without sacrificing performance. Our model serves as a pivotal technical advancement, facilitating timely cotton disease detection and providing valuable insights for the design of detection models for agricultural inspection robots and other resource-constrained agricultural devices.
... These datasets can be used for predictive PA yield analysis [58]. − Crop disease dataset: the dataset of image-based crop disease [59] and PlantVillage [60] are used for extracting disease information of soybeans, maize, and wheat. The dataset is labelled with a degree of disease severity. ...
... After the annotation is carried out, these datasets are required to be reused for further adoption in research work, and the existing system finds its sources from various publicly available platforms, e.g., Zenodo, Mendley Data, Open Science Framework, Harvard Dataverse, GitHub, Figshare, Dryad, and CyVerse. At present, there are a total of 34 image dataset that has been classified into 10 datasets for research on fruit detection, 15 dataset that are used for research on weed control, nine datasets for other different research areas of PA open image V4 dataset is another new dataset which is witnessed to be adopted by various research work as witnessed in study of Kuznetsova et al. [80], while PlantVillage (Mohanty et al. [59]) and Leafsnap (Redmon and Farhadi [81]) are another dataset which is seen to be frequently adopted in existing research work. Table 3 highlights the dataset characteristics increasingly adopted for research work in PA to address various issues. ...
Precision agriculture (PA) is meant to automate the complete agricultural processes with the sole target of enhanced crop yield with reduced cost of operation. However, deployment of PA in internet of things (IoT) based architecture demands solutions towards addressing various challenges where most are related to proper and precise predictive management of agricultural data. In this perspective, it is noted that learning-based approaches have made some contributory success towards addressing different variants of issues in PA; however, such methods suffer from certain loopholes, primarily related to the non-inclusion of practical constraints of IoT infrastructure in PA and lack of emphasis towards bridging the trade-off between higher accuracy and computational burden that is eventually associated with this. This paper contributes towards highlighting the strengths and weaknesses of recent learning approaches and contributes towards novel findings.
... Mohanty et al. [22], a dataset comprising 54,306 images of both diseased and healthy plant leaves, collected under controlled conditions from a public dataset, is employed. A deep convolutional neural network is trained to identify crop species and detect diseases or their absence. ...
An overview of methods for identifying plants diseases is given in this article. Each sample is categorized by being divided into various groups. The approach of classification involves identifying healthy and diseased leaves based on morphological traits including texture, color, shape, and pattern, among others. Sorting and categorizing plants can be challenging, especially when doing so across a large area, due to the closeness of their visual qualities. There are several methods based on computer vision and image processing. Selecting the right categorization method can be difficult because the outcomes rely on the data you supply. There are several applications for the categorization of plant leaf diseases in fields like agriculture and biological research. This article gives a summary of several approaches currently in use for identifying and categorizing leaf diseases, as well as their benefits and drawbacks. These approaches include preprocessing methods, feature extraction and selection methods, datasets employed, classifiers, and performance metrics
