Sanjiv Gupta

Artificial intelligence (AI) and deep learning (DL) for plant disease detection are emerging research areas. DL methods generally require a large amount of annotated data for training, which is often costly, time-consuming, and infeasible. This article addresses the data scarcity problem and proposes a few-shot learning (FSL) method for barley plant disease detection. To prepare a dataset, we captured images from outdoor test-bed trials (at two different growth stages of plants across multiple paddocks) under various weather conditions, such as sunny and cloudy. The images are divided into patches and manually labelled into five classes: no-disease, net form net blotch (NFNB) (which is classified into two stages: early and severe), spot form net blotch (SFNB), and scald. We name this as the Barley dataset. We also used the publicly available cassava dataset, which has five classes. The datasets are then applied to the proposed FSL pipeline, which only uses as few as five images for each class in training. We use the Swin transformer as the network backbone. The method with the Swin-B variant as the feature extractor achieved a detection accuracy of 91.80% and 97.93% on the barley disease and cassava datasets, respectively. The result indicates that our FSL model can efficiently perform and classify barley disease with small training data.
•A few-shot learning method is developed to address data scarcity problems.•Results on the collected plant disease data warrant the model’s potential.•Cutting-edge transformers, e.g. Swin-B, perform well given only five training images•Meta-training and transfer learning significantly improve performance.•Apparent disease symptoms can be detected and used in various applications.

Centuries of barley (Hordeum vulgare) domestication and selection has resulted in reduced genetic diversity in modern cultivars, limiting breeder's options to select desirable traits. Barley landraces, heirloom varieties, and wild relatives are substantially more variable and can be exploited to reintroduce favorable genes and alleles. Five doubled haploid populations were phenotyped for net form net blotch (NFNB) disease, caused by the pathogen Pyrenophora teres f. teres, at three growth stages. Major, moderate, and minor effect quantitative trait loci (QTLs) associated with NFNB resistance were detected on six of the seven barley chromosomes, with percentage of explained variance (PEV) ranging from less than 10% to over 70%. Previously established major (PEV > 50%) and moderate (PEV 10%–40%) effect QTLs on 3H and 6H were detected against the Australian isolate used, as well as moderate and minor QTLs (PEV < 10%) distributed on 2H, 3H, 4H, and 5H. Differences in effect sizes of individual QTL were apparent between growth stages, tapering up toward heading or down from seedlings, together with growth stage-specific and synergistic QTL. Several of these QTL represent novel sources of resistance that may be combined for durable NFNB resistance.

Powdery mildew (PM), caused by Blumeria hordei, can occur at all post emergent stages of barley and constantly threatens crop production. To identify more genes for effective resistance to powdery mildew for use in breeding programs, 696 barley accessions collected from different regions of the world were evaluated for PM resistance at seedling and adult growth stages in three different states of Australia. These barley accessions were genotyped using DArTSeq with over 18,000 markers for a genome-wide association study (GWAS). Using the FarmCPU model, 54 markers showed significant associations with PM resistance scored at the seedling and adult-plant stages in different states of Australia. Another 40 markers showed tentative associations (LOD > 4.0) with resistance. These markers are distributed across all seven barley chromosomes. Most of them were grouped into eleven QTL regions, coinciding with the locations of most of the reported resistance genes. Two major MTAs were identified on chromosome arms 3HS and 5HL, with one on 3HS contributing to adult plant resistance and the one on 5HL to both seedling and adult plant resistance. An MTA on 7HS contributed mainly to the adult-plant resistance, while another one on chromosome arm 1HS made a significant contribution to the seedling resistance.

Plant disease negatively impacts food production and quality. It is crucial to detect and recognise plant diseases correctly. Traditional approaches do not offer a rapid and comprehensive management system for detecting plant diseases. Deep learning techniques (DL) have achieved encouraging results in discriminating patterns and anomalies in visual samples. This ability provides an effective method to diagnose any plant disease symptoms automatically. However, one of the limitations of recent studies is that in-field disease detection is underexplored, so developing a model that performs well for in-field samples is necessary. The objective of this study is to develop and investigate DL techniques for in-field disease detection of barley (Hordeum vulgare L.), one of the main crops in Australia, given visual samples captured at barley trials using a consumer-grade RGB camera. Consequently, A dataset was captured from test-bed trials across multiple paddocks infected with three diseases: net form net blotch (NFNB), spot form net blotch (SFNB), and scald, in various weather conditions. The collected data, 312 images (6000 × 4000 pixels), are divided into patches of 448 × 448 pixels, which are manually annotated into four classes: no-disease, scald, NFNB and SFNB. Finally, the data was augmented using random rotation and flip to increase the dataset size. The generated barley disease dataset is then applied to several well-known pre-trained DL networks such as DenseNet, ResNet, InceptionV3, Xception, and MobileNet as the network backbone. Given limited data, these methods can be trained to detect anomalies in visual samples. The results show that MobileNet, Xception, and InceptionV3 performed well in barley disease detection. On the other hand, ResNet showed poor classification ability. Moreover, Augmenting the data improves the performance of DL networks, particularly for underperforming backbones like ResNet, and mitigates the limited data access for these data-intensive networks. The augmentation step improved MobileNet performance by approximately 6 %. MobileNet achieved the highest accuracy of 98.63 % (the average of the three diseases) in binary classification and an accuracy of 93.50 % in multi-class classification. Even though classifying SFNB and NFNB is challenging in the early stages, MobileNet achieved the minimum misclassification rate among the two diseases. The results show the efficiency of this model in diagnosing barley diseases using complex data collected from the field environment. In addition, the model is lighter and comprises fewer trainable parameters. Consequently, MobileNet is suitable for small training datasets, reducing data acquisition costs.
[Display omitted]

Spot form net blotch, caused by Pyrenophora teres f. maculata, is a significant global disease of barley (Hordeum vulgare). Baudin, a barley cultivar that was until recently extensively grown in Western Australia, was reported as having minor seedling resistance. However, Baudin was highly susceptible to a local isolate, M3, suggesting that this isolate had gained virulence against a major susceptibility gene. M3 causes atypical lesions with pale centers early in the infection, with initial screens of a segregating population indicating that this was determined by a single locus in the Baudin genome. The susceptibility was semidominant in F-1 progeny and the susceptibility gene, designated Spm1 (Susceptibility to P. teres f. maculata 1), mapped to a 190-kb section of the resistance gene-rich Mla region of chromosome 1H. Phenotyping with Ptm SP1, a non-M3 pathotype, identified a seedling resistance locus on 2H. Minor gene resistance is generally regarded as potentially durable, but our findings suggest the resistance to spot form net blotch in Baudin is nullified by strong susceptibility conferred by a separate locus on 1H.

Scald caused by Rhynchosporium commune is one of the most destructive barley diseases worldwide. Accumulation of adult plant resistance (APR) governed by multiple resistance alleles is predicted to be effective and long-lasting against a broad spectrum of pathotypes. However, the molecular mechanisms that control APR remain poorly understood. Here, quantitative trait loci (QTL) analysis of APR and fine mapping were performed on five barley populations derived from a common parent Vlamingh, which expresses APR to scald. Two QTLs, designated QSc.VR4 and QSc.BR7, were detected from a cross between Vlamingh and Buloke. Our data confirmed that QSc.VR4 is an effective and stable APR locus, residing on the short arm of chromosome 4H, and QSc.BR7 derived from Buloke may be an allele of reported Rrs2. High-resolution fine mapping revealed that QSc.VR4 is located in a 0.38 Mb genomic region between InDel markers 4H2282169 and 4H2665106. The gene annotation analysis and sequence comparison suggested that a gene cluster containing two adjacent multigene families encoding leucine-rich repeat receptor kinase-like proteins (LRR-RLKs) and germin-like proteins (GLPs), respectively, is likely contributing to scald resistance. Adult plant resistance (APR) governed by QSc.VR4 may confer partial levels of resistance to the fungus Rhynchosporium commune and, furthermore, be an important resource for gene pyramiding that may contribute broad-based and more durable resistance.

Barley yellow dwarf virus is a widespread disease affecting plant growth and yield in cereal crops including barley. Complete resistance to BYDV encoded by a single gene is lacking in barley. To identify novel resistance genes that can be further utilised in breeding for plant disease resistance, a doubled haploid population originated from a cultivated barley with a known resistance gene and a wild barley was constructed and assessed for barley yellow dwarf tolerance in three trials with two in Tasmania (TAS) and one in Western Australia (WA). We identified two Quantitative trait loci (QTL) in both Tasmanian trials, and four QTL in Western Australian trial. Two QTL from TAS trials were also detected from WA. The QTL on chromosome 3H corresponds to the known major resistance gene Ryd2. The other QTL, Qbyd-5H, represents a potential new resistance locus and contributed 7.0 similar to 10.4% of total phenotypic variation in the three trials. It was mapped within the interval of 125.76 similar to 139.24 cM of chromosome 5H. Two additional minor effect QTL were identified on chromosome 7H from WA trial, contributing slightly less effect on BYD tolerance. The consistently detected new gene on chromosome 5H will potentially serve as a novel source of tolerance to achieve more sustainable resistance to BYDV in barley.

Adult plant resistance against plant pathogens is of interest as a means to achieve durable resistance. Prior to this research, the barley lines CLE210 (from Uruguay) and Denar (from the Czech Republic) had been reported to exhibit adult-plant resistance against powdery mildew. Here, populations of doubled haploid lines from crosses of these lines with the susceptible cultivar Baudin were evaluated for powdery mildew resistance in field experiments. Using linkage maps constructed from genotyping-by-sequencing (GBS) data, it was determined that differences in resistance were largely attributable to a region on the long arm of chromosome 5H (5HL). Therefore, KASP™ assays were developed based on GBS tag sequences mapped on that chromosome, providing more reliable genetic maps. In each population, a large-effect QTL was mapped on 5HL. As no sequence variation was detected between CLE210 and Denar in this region of 5HL, the two sources of resistance may be identical by descent in the QTL region and carry the same resistance gene. Marker assays from the QTL region were evaluated on a panel of barley lines, providing information that breeders could use to select assays for use in marker-assisted selection.

Barley powdery mildew caused by Blumeria graminis f. sp. hordei is a most devastating disease in Western Australia and elsewhere. The disease has a negative impact on grain yield and quality and, thus, affects its marketability. The detection of pathogen variability has further impacted sustainable barley production. Various isolates have been reported to have broken down the resistance genes present in the upcoming barley cultivars. Because of this there is an increasing importance to identify novel resistances to powdery mildew with its effectiveness across barley growing regions of Australia. A set of 12 lines out of 101, which have originated from Australia, Czech Republic, South America, Sweden, South Africa and CIMMYT/ICARDA, was studied for growth stage responses in which 11 lines were identified to have adult plant resistance (seedling susceptibility). This set was evaluated over 2 years and across five different environments in Western Australia and Tasmania. Majority of the lines are effective across different environments and warrants its characterisation. The set holds a great promise and can broad narrow genetic base of current barley cultivars. This is a first report of identification of adult plant resistance for powdery mildew in barley lines from Australia.

Rhynchosporium secalis can overcome a single resistance gene of barley in a relatively short period of time. Novel genes and quantitative trait locis (QTLs) are therefore vital to control scald in barley. A population of 220 double haploid lines was developed from a cross of Vlamingh and WABAR2147, where Vlamingh showed adult plant resistance (APR) and WABAR2147 showed seedling resistance to a group of isolates. The population was tested for APR to scald under natural infection in two consecutive seasons in addition to a seedling screen with three isolates. One single gene was mapped to chromosome 6H based on the seedling test, and two QTLs (QSc.VlWa.4H and QSc.VlWa.6H) were mapped to chromosomes 4H and 6H based on APR. Epistatic interaction was observed between the two QTLs, and environment/QTL interaction was only observed for QSc.VlWa.6H which co-segregated with the seedling resistance gene and contributed to basal resistance against scald during whole growth stages. QSc.VlWa.4H explained 42.5 and 57.8 % of the phenotypic variation in the two independent trials when the effect of QSc.VlWa.6H was excluded from the analysis. We developed a high-density consensus genetic map with 7,876 molecular makers and anchored 43 QTLs and 7 genes for scald resistance from different mapping populations. No known QTLs or genes were reported in a similar position to QSc.VlWa.4H, and it was the first major QTL for APR of scald on chromosome 4HS in barley. Combination of the two QTLs achieved better and stable scald resistance across four different environments.