Wheat Leaf Diseases Research Articles

Automated Assessment of Wheat Leaf Disease Spore Concentration Using a Smart Microscopy Scanning System

An advanced approach to the automated assessment of a microscopic slide containing spores is presented. The objective is to develop an intelligent system for the rapid and precise estimation of phytopathogenic spore concentration on microscopic slides, thereby enabling automated processing. The smart microscopy scanning system comprises an electronic microscope, a coordinate table, and software for the control of the coordinate table and image processing. The developed smart microscopy scanning system processes the entire microscope slide with multiple exposed strips, which are automatically determined based on the novel two-stage algorithm. The analysis of trained convolutional neural networks employed for the detection of spore phytopathogens demonstrates high precision and recall metrics. The system is capable of identifying and counting the number of spores of phytopathogenic fungi species Blumeria graminis, Puccinia striiformis, and Pyrenophora tritici-repentis on each exposed strip. A methodology for estimating the spore distribution on a microscopic slide is proposed, which involves calculating the average spore concentration density.

First report of wheat leaf spot disease caused by Curvularia inaequalis in China.

Wheat (Triticum aestivum) is an economically important crop widely cultivated in China. In August 2022, brown oval leaf spots with yellow halos were observed on approximately 10% wheat seedlings over an area of about 1 hectare in Xining City, Qinghai Province, which adversely affected wheat growth and production. Six diseased leaves were collected from the field in Huangyuan county (101°69' E, 37°04' N). The 0.5 cm × 0.5 cm pieces were cut from the border between healthy and diseased regions of the sampled leaves, surface sterilized for 10 s in 75% ethanol, followed by a 1% NaClO for 90 s, and rinsed three times with distilled sterile water. The pieces of leaf tissue were dried with sterile tissue, and plated on potato dextrose agar (PDA) amended with streptomycin (0.02 g/L) and ampicillin sulfate (0.05 g/L) to eliminate bacterial contamination. The dishes were placed in an incubator at 25°C for 72 h in dark. Three isolates, WGC201, WGC202 and WGC203, were obtained by a single-spore culture method. Fungal colonies on PDA media were dark green (Fig. 1A and 1B). Conidiophores were septate and geniculate terminals, while conidia exhibited straight or slightly curved forms with four transverse septa, the central cell being notably longer and wider than the others. The size of such conidia were 27.34 µm to 40.62 µm× 11.61 µm to 15.97 µm (number = 50) (av. 32.71 µm× 13.11 µm) (Fig. 1C and 1D) (Moubasher et al. 2010). The internal transcribed spacer (ITS) region of nuclear ribosomal DNA and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene were amplified and sequenced using universal primers ITS1/ITS4 and GPDF/GPDR (White et al. 1990; Berbee et al. 1999). DNA sequences were deposited into the NCBI database (ITS, PP789629, PP801333, and PP801574; GAPDH, PP849124, PP849125, and PP849126). Phylogenetic analysis with a neighbor-joining method based on the concatenated sequences of ITS and GAPDH genes showed that the three isolates clustered within a C. inaequalis branch (Fig. 2). Based on morphological and molecular identification, the fungal isolates were identified as C. inaequalis. The pathogenicity test was conducted in a greenhouse at 25°C using a spore suspension method and three isolates were used. Conidia were produced on PDA media (25℃) for 14 days. Plates were washed with sterilized distilled water and filtered with cheese cloth. Conidial suspension was adjusted to a concentration of 1×107 conidia/mL. Fifteen healthy seedlings of a wheat cultivar Xiaoyan-6 at a 3-4 leaf stage were inoculated by evenly spraying a 100mL spore suspension. Plants inoculated with sterile water served as a control. All plants were covered with plastic bags for 3 days. At 7 days after inoculation, all pathogen-inoculated plants showed similar symptoms (brown leaf oval spots with yellow halos) with those observed in the field, while all plants inoculated with sterile water showed no symptoms (Fig. 1E and 1F). The pathogen was reisolated from the symptomatic leaves and proved to be C. inaequalis. Morphological, molecular and pathogenic results indicated that C. inaequalis is the pathogen causing wheat leaf disease in China. The results are consistent with a previous report in Azerbaijan (Özer et al. 2020). To our knowledge, this is the first report of C. inaequalis causing spot disease on wheat in China. The occurrence, spread and economic importance to different wheat cultivars of the emerging disease in China will be further investigated and evaluated.

Search for sources of resistance to leaf diseases among present winter bread wheat varieties in the southern part of the Rostov region

The purpose of the current work was to search for sources of winter wheat varieties’ resistance to the most common pathogens of the area under artificial infectious backgrounds and recommend them for inclusion in the breeding process. The study was carried out on the experimental plots of the laboratory for plant immunity and protection of the FSBSI Agricultural Research Center “Donskoy” in 2021–2023. In order to expand the genetic diversity in immunity of newly developed varieties, it is necessary to use new sources of resistance. The objects of study were 78 winter wheat varieties from interstation variety testing, representing various breeding institutions in Russia and some foreign varieties. The study material was presented by the North Caucasian populations of pathogens of such wheat leaf diseases as powdery mildew, brown and yellow rust, and leaf spot (Septoria). Infectious backgrounds were formed according to generally accepted methods, using spore material of pathogens both stored and collected from crops in the previous year (types of rust) and collected from overwintered plants in the spring (powdery mildew, leaf spot). Weather conditions varied across the years of study (mainly in the autumn), but the damage to susceptible test varieties in the experiments was maximum. As a result of the study of the varieties on their resistance to one pathogen, there have been identified 23 varieties resistant to powdery mildew, 56 ones to leaf rust, 47 ones to yellow rust, 8 varieties to leaf spot. There has been given a characteristic of varieties according to various degree of resistance or susceptibility to each pathogen. On the resistance to two pathogens, there have been identified 17 varieties resistant to brown and yellow rust, 2 varieties resistant to brown rust and powdery mildew, 2 varieties resistant to yellow rust and powdery mildew, 1 variety resistant to brown rust and leaf spot. Ten varieties have shown resistance to no one pathogen, but they have had moderate resistance or moderate susceptibility to several other pathogens. 20 varieties were resistant to three diseases in various combinations, and 16 of them were resistant to brown, yellow rust, and powdery mildew. One variety ‘MV Nador’ from Hungary showed resistance to all 4 pathogens. All resistant varieties identified in the study can replenish the supply of winter wheat sources to the complex of leaf diseases for breeding purposes or be used in integrated crop protection against leaf diseases.

Wheat Leaf Disease Detection: A Lightweight Approach with Shallow CNN Based Feature Refinement

Improving agricultural productivity is essential due to rapid population growth, making early detection of crop diseases crucial. Although deep learning shows promise in smart agriculture, practical applications for identifying wheat diseases in complex backgrounds are limited. In this paper, we propose CropNet, a hybrid method that utilizes Red, Green, and Blue (RGB) imaging and a transfer learning approach combined with shallow convolutional neural networks (CNN) for further feature refinement. To develop our customized model, we conducted an extensive search for the optimal deep learning architecture. Our approach involves freezing the pre-trained model for feature extraction and adding a custom trainable CNN layer. Unlike traditional transfer learning, which typically uses trainable dense layers, our method integrates a trainable CNN, deepening the architecture. We argue that pre-trained features in transfer learning are better suited for a custom shallow CNN followed by a fully connected layer, rather than being fed directly into fully connected layers. We tested various architectures for pre-trained models including EfficientNetB0 and B2, DenseNet, ResNet50, MobileNetV2, MobileNetV3-Small, and Inceptionv3. Our approach combines the strengths of pre-trained models with the flexibility of custom architecture design, offering efficiency, effective feature extraction, customization options, reduced overfitting, and differential learning rates. It distinguishes itself from classical transfer learning techniques, which typically fine-tune the entire pre-trained network. Our aim is to provide a lightweight model suitable for resource-constrained environments, capable of delivering outstanding results. CropNet achieved 99.80% accuracy in wheat disease detection with reduced training time and computational cost. This efficient performance makes CropNet promising for practical implementation in resource-constrained agricultural settings, benefiting farmers and enhancing production.

Disease Detection In Rice And Wheat Leaves: A Comparative Study On Various Deep Learning Techniques

Rice and wheat are considered as the most significant foods in agriculture all over the globe. Around 50% of all calories consumed in the human diet are rendered by both rice and wheat. Rice is a rich source of carbohydrates that is the main energy source of the human body. Wheat, which comprises vitamins and minerals, is a staple food source. Also, wheat is extensively used as flour to make a variety of food products. However, the disease that occurs in the leaves of rice and wheat could decelerate the production of these two food sources. Thus, timely detection of disease that occurs within the rice and wheat leaves is very significant. To detect Rice Leaf Diseases (RLD) and Wheat Leaf Diseases (WLD), numerous conventional methods are established. Specifically, You Only Look Once (YOLO), Faster Regionbased Convolutional Neural Network (FRCNN), and Single Shot Detector (SSD) are extensively implemented in various works to detect diverse leaf diseases in rice and wheat plants. Therefore, in this review, the merits and demerits of the traditional Object Detection (OD) models in RLD and WLD detection are provided systematically. The robustness of the Deep learning (DL) techniques in detecting various kinds of leaf diseases in rice and wheat plants with a classification accuracy of 99% and a precision of 98% is proved by the analysis outcomes.

Winter Durum Wheat Disease Severity Detection with Field Spectroscopy in Phenotyping Experiment at Leaf and Canopy Level

Accurate disease severity assessment is critical for plant breeders, as it directly impacts crop yield. While hyperspectral remote sensing has shown promise for disease severity assessment in breeding experiments, most studies have focused on either leaf or canopy levels, neglecting the valuable insights gained from a combined approach. Moreover, many studies have centered on experiments involving a single disease and a few genotypes. However, this approach needs to accurately represent the challenges encountered in field conditions, where multiple diseases could occur simultaneously. To address these gaps, our current study analyses a combination of diseases, yellow rust, brown rust, and yellow leaf spots, collectively evaluated as the percentage of the diseased leaf area relative to the total leaf area (DA) at both leaf and canopy levels, using hyperspectral data from an ASD field spectrometer. We quantitatively estimate overall disease severity across fifty-two winter durum wheat genotypes categorized into early (medium milk) and late (late milk) groups based on the phenophase. Chlorophyll content (CC) within each group is studied concerning infection response, and a correlation analysis is conducted for each group with nine vegetation indices (VI) known for their sensitivity to rust and leaf spot infection in wheat. Subsequent parametric (linear and polynomial) and nonparametric (partial least squares and kernel ridge) regression analyses were performed using all available spectral bands. We found a significant reduction in Leaf CC (>30%) in the late group and Canopy CC (<10%) for both groups. YROI and LRDSI_1 are the VIs that exhibited notable and strong negative correlations with Leaf CC in the late group, with a Pearson coefficient of −0.73 and −0.72, respectively. Interestingly, spectral signatures between the early and late disease groups at both leaf and canopy levels exhibit opposite trends. The regression analysis showed we could retrieve leaf CC only for the late group, with R2 of 0.63 and 0.42 for the cross-validation and test datasets, respectively. Canopy CC retrieval required separate models for each group: the late group achieved R2 of 0.61 and 0.37 (cross-validation and test), while the early group achieved R2 of 0.48 and 0.50. Similar trends were observed for canopy DA, with separate models for early and late groups achieving comparable R2 values of 0.53 and 0.51 (cross-validation) and 0.35 and 0.36 (test), respectively. All of our models had medium accuracy and tended to overfit. In this study, we analyzed the spectral response mechanism associated with durum wheat diseases, offering a novel crop disease severity assessment approach. Additionally, our findings serve as a foundation for detecting resistant wheat varieties, which is the most economical and environmentally friendly management strategy for wheat leaf diseases on a large scale in the future.

An Interpretable Deep Learning for Early Detection and Diagnosis of Wheat Leaf Diseases

Wheat is a strategic crop that ensures food security and sustainability. Due to climate change, Egypt faces many challenges in the context of wheat disease. This led the government to set goals to achieve sustainable development, represented by Egypt’s Vision 2030. Automated systems using artificial intelligence can help in the early identification of wheat diseases and reduce effort and time as well. In this paper, we introduce a novel deep learning model (Mobile-DNN-Net) to identify wheat diseases. The mobile-DNN-Net model is evaluated using a wheat disease dataset between 15 different classes. Our model (Mobile-DNN-Net) is a hybrid between MobileNet and Deep convolution neural network (DCNN). Also, Grad-Cam techniques into a convolutional neural network (CNN) defect detection model to enhance its transparency and comprehensibility. Grad-CAM precisely identifies the precise regions of the input image that exert the greatest influence on the model's prediction, thereby enhancing the clarity and comprehension of the detection process. The mobile-DNN-Net model is compared to other DL models such as Xception, MobileNet, InceptionV3, and VGG19. The proposed model shows superior results compared to other models.

Improving Wheat Leaf Disease Image Classification with Point Rend Segmentation Technique

Improving Wheat Leaf Disease Image Classification with Point Rend Segmentation Technique

A robust and light-weight transfer learning-based architecture for accurate detection of leaf diseases across multiple plants using less amount of images.

Leaf diseases are a global threat to crop production and food preservation. Detecting these diseases is crucial for effective management. We introduce LeafDoc-Net, a robust, lightweight transfer-learning architecture for accurately detecting leaf diseases across multiple plant species, even with limited image data. Our approach concatenates two pre-trained image classification deep learning-based models, DenseNet121 and MobileNetV2. We enhance DenseNet121 with an attention-based transition mechanism and global average pooling layers, while MobileNetV2 benefits from adding an attention module and global average pooling layers. We deepen the architecture with extra-dense layers featuring swish activation and batch normalization layers, resulting in a more robust and accurate model for diagnosing leaf-related plant diseases. LeafDoc-Net is evaluated on two distinct datasets, focused on cassava and wheat leaf diseases, demonstrating superior performance compared to existing models in accuracy, precision, recall, and AUC metrics. To gain deeper insights into the model's performance, we utilize Grad-CAM++.

Wheat leaf disease classification using modified ResNet50 convolutional neural network model

Wheat leaf disease classification using modified ResNet50 convolutional neural network model

A Comprehensive Approach Toward Wheat Leaf Disease Identification Leveraging Transformer Models and Federated Learning

A Comprehensive Approach Toward Wheat Leaf Disease Identification Leveraging Transformer Models and Federated Learning

Wheat leaf diseases classification and severity analysis using HT-CNN and Hex-D-VCC-based boundary tracing mechanism.

Wheat is one among the significant crops for humans. Significant fungal illnesses of wheat are brought on by multiple pathogens. Wheat output could be enhanced by the early identification of wheat leaf disease. Thus, a novel hyperparameter tanh-based convolutional neural network (HT-CNN)-based wheat leaf disease prediction is proposed with its severity level. Here, initially, the red, green, and blue (RGB) images are converted into a hue saturation value (HSV) image. Next, the small probability space filtering is applied to the V component. Afterward, the contrast of the V component has been enhanced. The obtained HSV image is converted into the RGB image. Then, by employing weighted Canberra distance-based K-means (WCD-K means), the affected and normal regions are segmented. Next, the image is binarized. Afterward, for tracing a boundary around disease-affected region, the hex directional vertex chain code (Hex-D-VCC) is applied over the binarized image, and then the features are extracted. By employing baker's map-based Harris hawks optimization (BM-HHO), the optimal features are selected. For classifying disease, the selected features are further given into the HT-CNN, and the severity level is calculated to minimize the yield loss. As per the experimental result, the proposed model shows higher accuracy and efficacy when analogized to the other methods.

Recognition of Wheat Leaf Diseases Using Lightweight Convolutional Neural Networks against Complex Backgrounds.

Wheat leaf diseases are considered to be the foremost threat to wheat yield. In the realm of crop disease detection, convolutional neural networks (CNNs) have emerged as important tools. The training strategy and the initial learning rate are key factors that impact the performance and training speed of the model in CNNs. This study employed six training strategies, including Adam, SGD, Adam + StepLR, SGD + StepLR, Warm-up + Cosine annealing + SGD, Warm-up + Cosine, and annealing + Adam, with three initial learning rates (0.05, 0.01, and 0.001). Using the wheat stripe rust, wheat powdery mildew, and healthy wheat datasets, five lightweight CNN models, namely MobileNetV3, ShuffleNetV2, GhostNet, MnasNet, and EfficientNetV2, were evaluated. The results showed that upon combining the SGD + StepLR with the initial learning rate of 0.001, the MnasNet obtained the highest recognition accuracy of 98.65%. The accuracy increased by 1.1% as compared to that obtained with the training strategy with a fixed learning rate, and the size of the parameters was only 19.09 M. The above results indicated that the MnasNet was appropriate for porting to the mobile terminal and efficient for automatically identifying wheat leaf diseases.

Classification of Image-Based Wheat Leaf Diseases using Deep Learning Approach: A Survey

This paper focuses on detecting leaf diseases in wheat plants from the beginning to the end of the plant's life cycle. It highlights the best techniques for detecting various types of wheat leaf diseases and emphasizes the use of computer vision, image processing, and machine learning. The main focus is on classifying these diseases through deep convolutional neural networks, a popular image recognition and classification approach. The paper reviews various techniques for classifying image-based wheat leaf diseases, including spot blotch, stripe rust, brown rust, and powdery mildew. The paper aims to summarize the state-of-the-art techniques for detecting wheat leaf diseases.

Endogenous Salicylic acid Estimation in Wheat leaves treated with Salicylic acid and infected with Alternaria triticina

This study was aime to measure the accumulation of endogenous salicylic acid, as indicator for systemic acquired re-sistance of wheat plants of “Utique” variety. The plant was sprayed with salicylic acid (SA) or water as control, at the five-leaf stage, later infected with leaf blight caused by Alternaria triticina. Leaf samples were removed after 10, 20, and 30 days of inoculation to test their endogenous content of salicylic acid as it is the primary internal signal indicating the emergence of systemic acquired resistance in plants, by the spectrophotometer measuring. Endogenous SA values were ascending dramatically from 10 days to 20 days and maximum with 30th day significantly, while control plants exhibit lower values in all day periods, insignificantly. SA treatments proved reduction in disease incidence after 10 days with 54% and after 20 days with 64%, while after 30 days the reduction recorded high percentage of 80%. In comparison between the time intervals, disease severity was clearly reduction and reached to 83% after 10 days of inoculation then decreased to 72% for both 20th and maintained its stability on the 30th day of inoculation by 72%. This study was proved the reduction of wheat leaf blight incidence and disease severity as a result of treatment by using 1Millimolar (mM) of SA was leading to accumulation of significant levels of endogenous SA as indicator for internal induced resistance in plant.

Soil drenching and foliar spray with bioagents for reducing wheat leaf diseases under natural field conditions

Foliar diseases, like powdery mildew, septoria leaf blotch (SLB) and stem rust, affect wheat plants causing severe yield losses. Biofertilizers Pseudomonas fluorescens Migula (NRC2041) and Bacillus subtilis Ehrenberg (NRC313) or the arbuscular mycorrhizal fungi Glomus mosseae Tul. & C. Tul. (NRC212A) and G. fasciculatum Tul. & C. Tul. (NRC212B) were used for soil treatment followed by foliar spray with Trichoderma harzianum or P. fluorescens to manage severity of wheat foliar diseases under field conditions for two successive growing seasons. Under field conditions, at all soil drench and foliar application treatments, Sids 14 was the most sensitive cultivar for high disease severity followed by Misr 2 and Giza 168. Also, the obtained results showed that the most effective treatments were soil drench with mycorrhizae followed by foliar spray with either T. harzianum or P. fluorescens. Meanwhile, the fungicide Amistar showed a superior reduction in disease severity compared to other applied treatments. In addition, the increased yield was correlated to the decreased severity of the recorded diseases. The present study suggests using bioagents for reducing the severity of wheat leaf diseases and increasing the grain yield under field conditions.

Wheat leaf disease identification based on deep learning algorithms

Wheat is one of the most important crops and food sources in the world. However, wheat leaf diseases have a major impact on growth. An accurate diagnosis of wheat leaf diseases is essential for wheat quality and the agricultural economy. To improve the identification precision of wheat leaf diseases, we propose an integrated deep learning algorithm, which combines a residual channel attention block (RCAB), a feedback block (FB), elliptic metric learning (EML), and a convolutional neural network (CNN) and call it RFE-CNN. First, we utilized two parallel CNNs to extract the basic features of healthy and diseased wheat leaves, respectively. Second, we used residual channel attention blocks to optimize the basic features. Third, we used feedback blocks to train the previous features. Finally, we sent these features into a CNN and elliptic metric learning for processing and classification. The experimental results demonstrate that the proposed model is superior to VGG-19, ZFNet, GoogLeNet, Inception-V4, and Efficient-B7 in some aspects, such as shorter time consumption, higher recognition precision, and stronger adaptive ability. The overall classification accuracy was 98.83%, and the maximum testing accuracy was 99.95%. We obtained an average accuracy score of 99.50% on the open-source databases viz., CGIAR, Plant Diseases, LWDCD 2020, and Plant Pathology. The proposed method has a good reference for the promotion of intelligent crop disease and insect pest detection. The recognition rate is relatively low for the samples of different ecological locations and wheat varieties. Therefore, our algorithms need to be further improved to achieve a better balance. We will use hyperspectral imaging technology to obtain more spectral data on wheat leaf diseases and send them into deep learning models for classification research.

Wheat Disease Detection for Yield Management Using IoT and Deep Learning Techniques

Our economy is mostly based on agriculture. One of the difficult problems in the agriculture sector is crop yield predictions. Crop yield prediction using a machine learning algorithm with the help of IoT increases the production of wheat yield and improves the quality of yield. Today's low agricultural production is a problem for farmers. Low crop output is mainly caused by a lack of information regarding soil fertility and crop selection, and proper crop selection is the key to maximizing crop yield. One of the interesting agricultural research areas where deep learning (DL) algorithm concepts can be used is the identification of wheat disease from images. We consider two leaf diseases septoria and stripe rust and also take a healthy leaf and then do a comparison between the leaves using CNN. As a contribution, we developed a system ML with a neural network mobilenet and efficient net-b3 that detects wheat leaf disease and improves accuracy gradually. Moreover, we do a complete review of yield management in which IoT sensors are used with machine learning algorithms. This study aims to create a system that can correctly choose a crop for maximum yield utilizing IoT devices and machine learning (ML) algorithms. We achieve 97% accuracy using mobilenet which is better than the efficient net. The presented work also applied different image augmentation techniques to remove the problem of overfitting. The presented work is compared with the state-of-the-art method in terms of accuracy and precision score.

Hybrid Deep Learning Model to Detect Uncertain Diseases in Wheat Leaves

To improve the wheat crop’s yield, leaf disease detection has been considered as an important research area. In the digital image processing field, computer vision and deep learning models have gained better performance. Several deep learning-based techniques have also been used previously for wheat leaf disease detection, but the accuracy is still considered a challenging task. The basic convolutional neural networks (CNNs) model of deep learning has poor performance for abnormal image orientation, rotation, etc. Hence, to overcome this issue, a hybrid model that is a combination of VGG-16 and a capsule network is implemented in this paper. VGG-16 is a pre-trained model used for the recognition and categorization of objects. CapsNet is a novel architecture for deep learning design to solve CNN’s limitations for rotation, tillage, or other picture anomalies. Thus, a hybrid model has been developed, integrating VGG-16 and capsule network characteristics. The new hybrid model is termed here as WheCNet. Hence, the VGG-16 and capsule network features are combined into a new hybrid model called WheCNet. Features are first extracted by using the VGG-16 model. When there are misalignment issues with the existing deep learning models, CapsNet layers are used. Finally, sigmoid activation functions and fully connected layers are employed. Overfitting can be avoided by introducing dropouts, which are also used to build more generalized WheCNet models. The performance of WheCNet is compared with ResNet, MobileNet, XceptionNet, and VGG for the same dataset. After the successful implementation, WheCNet exhibits validation accuracy of 98%, while ResNet, MobileNet, XceptionNet and VGG achieve accuracy of 65%, 81%, 52%, and 93%, respectively. Hence, the proposed hybrid model, WheCNet will simplify the detection of uncertain diseases in wheat leaves with higher accuracy.

Evaluation of Diverse Convolutional Neural Networks and Training Strategies for Wheat Leaf Disease Identification with Field-Acquired Photographs

Tools for robust identification of crop diseases are crucial for timely intervention by farmers to minimize yield losses. Visual diagnosis of crop diseases is time-consuming and laborious, and has become increasingly unsuitable for the needs of modern agricultural production. Recently, deep convolutional neural networks (CNNs) have been used for crop disease diagnosis due to their rapidly improving accuracy in labeling images. However, previous CNN studies have mostly used images of single leaves photographed under controlled conditions, which limits operational field use. In addition, the wide variety of available CNNs and training options raises important questions regarding optimal methods of implementation of CNNs for disease diagnosis. Here, we present an assessment of seven typical CNNs (VGG-16, Inception-v3, ResNet-50, DenseNet-121, EfficentNet-B6, ShuffleNet-v2 and MobileNetV3) based on different training strategies for the identification of wheat main leaf diseases (powdery mildew, leaf rust and stripe rust) using field images. We developed a Field-based Wheat Diseases Images (FWDI) dataset of field-acquired images to supplement the public PlantVillage dataset of individual leaves imaged under controlled conditions. We found that a transfer-learning method employing retuning of all parameters produced the highest accuracy for all CNNs. Based on this training strategy, Inception-v3 achieved the highest identification accuracy of 92.5% on the test dataset. While lightweight CNN models (e.g., ShuffleNet-v2 and MobileNetV3) had shorter processing times (<0.007 s per image) and smaller memory requirements for the model parameters (<20 MB), their accuracy was relatively low (~87%). In addition to the role of CNN architecture in controlling overall accuracy, environmental effects (e.g., residual water stains on healthy leaves) were found to cause misclassifications in the field images. Moreover, the small size of some target symptoms and the similarity of symptoms between some different diseases further reduced the accuracy. Overall, the study provides insight into the collective effects of model architecture, training strategies and input datasets on the performance of CNNs, providing guidance for robust CNN design for timely and accurate crop disease diagnosis in a real-world environment.