The application of animal manure can reshape microbial functions in soil-plant systems, but it remains unclear whether these responses are consistent across soil and plant compartments and whether manure-associated changes in root exudation provide a plausible indirect link to rhizosphere functions. Here, we investigated microbial functional profiles in different lettuce-associated compartments (episphere, endosphere, and root) and soils (surface soil and rhizosphere soil) after beef cattle manure application, and further characterized root exudates by untargeted GC-MS metabolomics. Shotgun metagenomics showed that carbon and amino acid metabolism were dominant functional categories across the soil-plant ecosystem. Pathway-wide differential abundance analysis (ANOVA with FDR correction) of all 131 detected KEGG pathways revealed a striking compartment-dependent pattern: no pathway reached FDR significance in any plant-associated compartment (episphere, endosphere, or root) following manure application, whereas 72 pathways were differentially abundant at FDR < 0.1 in surface soil. Beta-diversity analysis indicated that manure significantly affected overall soil microbial functional profiles, whereas plant-associated functional beta-diversity was primarily structured by compartment identity rather than treatment. Metabolomics identified 67 root-exudate compounds with retention-index similarities >95%, and several amino acids and organic acids showed higher peak areas under manure treatment. Because these compounds participate in microbial amino acid and carbon metabolism, they may have contributed to the pathway-level shifts observed in the rhizosphere; however, the present data do not establish a direct statistical linkage between exudate composition and microbial functional change. These results highlight strong compartment filtering in lettuce-associated microbiomes, with manure effects being buffered in plant tissues while driving broad functional shifts in surface soil.

The application of animal manure affects the composition of soil and plant-associated microbiome. The objective of this study was to evaluate the microbial functions in different compartments of lettuce (i.e., episphere, endosphere and root) and soil (i.e., surface soil and rhizosphere soil) in response to manure application. Shotgun metagenomic results show that carbon metabolism and amino acid metabolism are the primary metabolic pathways of the microbiomes in the soil-plant ecosystem. Manure application had significant (Adonis, p<0.05) effects on the soil (rhizosphere soil and surface soil) but not on (Adonis, p>0.05) plant (episphere, endosphere, and root) microbial functions. Microbial functional profiles of plant microbiome are more likely to be driven by plant compartments. Metabolomics results show that 67 compounds of root exudates with Retention Index similarities > 95% were found in the Gas Chromatography-Mass Spectrometry untargeted assay. Compounds with elevated peak area including citric acid (isocitric acid), amino acids (glutamic acid, lysine, norvaline, ornithine) were found in root exudate of lettuce planted in soil with manure application. These affected root exudates are involved in numerous metabolic pathways (especially amino acid metabolisms), suggesting that cattle manure may indirectly affect the metabolic function of rhizosphere microbiomes by altering root exudates. Our discoveries help to understand how organic fertilizer shapes the plant-microbiome interaction and provide insights into the enhanced construction of functional microbial systems in the agricultural setting.

Hyperspectral imaging has been used to determine plant stress status. However, the biological interpretation of the spectral changes remain less explored. This can be addressed by building associations between stress-induced biochemical changes and variations in spectral reflectance. To this end, we tested spectral response of sorghum brown midrib (bmr) mutants under varying water stress levels using hyperspectral imaging (650-1650 nm). The bmr mutants have reduced lignin concentrations in their vegetative tissue which was reflected as spectral differences. Under water stress, the spectral signatures diverged more between the wildtype and mutants compared to control conditions. The genotype-dependent variation in spectral trends under water limitation was associated with differential sensitivity of the genotypes to water-limitation induced changes in energy density. We show that the energy density and relative water content of the plant tissue can be estimated accurately from spectral reflectance. To reduce the computational load, LASSO was used to obtain 22 wavelengths across the camera spectral range (650-1650 nm) in dried samples, to accurately predict energy density comparable to PLSR estimates. The reported wavelengths represent a useful screening tool for fast and reliable calorimetric estimations in bioenergy breeding programs.

The land application of livestock manure can have significant effects on the emergence and proliferation of antibiotic resistance genes (ARGs) and antibiotic residues in the soil-plant ecosystem. While previous studies have documented the effects of manure application on ARGs in either soil or plant compartments, research addressing its impact on ARGs concurrently in both soil and plants remains relatively limited. This study aims to assess the occurrence of ARGs in and on lettuce cultivated in soil with beef cattle manure application. Results showed that seven ARGs (bla(TEM), erm(B), erm(F), tet(M), tet(O), tet(Q), and tet(X)) and one class 1 integron-integrase gene (intI1) were quantified in both soil and lettuce compartments following manure application. The relative abundance of manure-borne tet(M), tet(Q), and tet(X) was significantly elevated in surface soils (Kruskal-Wallis tests, p < 0.05). Notably, tet genes increased by 1-3 orders of magnitude within the lettuce endosphere and roots, revealing a potential transmission from soil to lettuce. In contrast, the relative abundance of bla(TEM), erm(B), and erm(F) increased only in the episphere and root but not within the endosphere of lettuce. Additionally, concentrations of tylosin in surface soil showed positive correlations with tet genes, suggesting their potential roles in facilitating the ARG proliferation in the soil-plant ecosystem. In summary, this study demonstrates that manure application promoted the transmission of manure-borne ARGs from soil to lettuce, highlighting a potential pathway for human exposure to antimicrobial resistance through the food chain. This finding underscores the need for the development of manure management practices to mitigate ARG spread in agriculture.

The application of Fourier transform infrared (FTIR) spectroscopy for non-structural carbohydrates (NSC) prediction as a tool for pre-breeding screening has immense potential but remains to be unexplored, because of technical challenges associated with these measurements. This study investigated the potential of employing FTIR spectroscopy as a high-throughput tool for forecasting NSC content, including total soluble sugar (TSS) and starch content, of 30 rice accessions from the Rice Diversity Panel 1 (RDP1) germplasm and RiceTec hybrids grown in 2019 (320 genotypes) and 2020 cropping (312 genotypes). Partial Least Squares (PLS) regression analysis was used to construct predictive models to estimate NSC content in flag leaves and stem of rice exposed to elevated and ambient nighttime air temperature during the flowering stage of rice. The TSS model exhibited a coefficient of determination (R2) value of 0.63 and root mean square error of prediction (RMSEP) values of 3.62 mg g− 1. Notably, the NSC model demonstrated a superior metric performance, with R2 = 0.66 and RMSEP of 5.58 mg g− 1. The predictive model created in this research effectively measured the NSC composition present in the flag leaves of rice. Expanding the sample size and incorporating additional principal components may enhance the model’s predictive accuracy. The FTIR technique can produce fast accurate results and resolve the high analytical costs. Overall, the use of FTIR in conjunction with PLS regression analysis provides a potential tool to advance our understanding of various rice genotypes, particularly concerning their ability to withstand abiotic stress such as HNT.

Rice is consumed by ∼50% of the global population, grown primarily in flooded paddy fields, and is susceptible to arsenic accumulation. Inorganic arsenic, particularly in reduced form (As(III)), is considered the most toxic and is more likely to accumulate in rice grains under flooded systems. We postulate that increased levels of highly reactive iron minerals, such as ferrihydrite, in paddy soils can regulate the bioavailability of arsenic and reduce its uptake by priming iron plaque formation. To clarify, two rice varieties, Norin and Sabharaj, differing in arsenic uptake rate, were grown in paddy soil under flooded conditions with arsenate (As(V)) spiked-irrigation water. 2-line ferrihydrite was added at 0.00% (control), 0.05%, and 0.10% w/w and served as the highly reactive iron species. Irrespective of rice varieties, total inorganic arsenic (As(III) + As(V)) in grains in ferrihydrite systems decreased by 85 to 93% compared to the control. These results support ferrihydrite's intrinsic role in controlling paddy soils' rhizosphere chemistry. Our findings indicate that fresh reactive iron minerals are critical in the early formation of root iron plaque, which enhances the defense mechanism against arsenic. The findings may have implications for reducing toxic inorganic arsenic accumulation in lowland rice.

Rising temperatures due to global warming can negatively impact rice grain quality and yield. This study investigates the effects of increased warmer night temperatures (WNT), a consequence of global warming, on the quality of rice kernel, particularly grain chalkiness. By integrating computational and experimental approaches, we used a rice grain metabolic network to discover the metabolic factors of chalkiness. For this, we reconstructed the rice grain genome-scale metabolic model (GSM), iOSA3474-G and incorporated transcriptomics data from three different times of the day (dawn, dawn 7 h, and dusk) for both control and WNT conditions with iOSA3474-G. Three distinct growth phases: anoxia, normoxia, and hyperoxia, were identified in rice kernels from the GSMs, highlighting the grain-filling pattern under varying oxygen levels. We predicted excess flux through histidine contributing to the biomass as a marker of normoxia, during which kernel chalkiness occurs. Moreover, similarly, we proposed tyrosine as a marker for the hyperoxic growth phase. We also proposed a potential link between monodehydroascorbate reductase, an enzyme with evolutionary significance dating back to the carboniferous era, in regulating the hyperoxic growth phase. Metabolic bottleneck analysis identified nucleoside diphosphate kinase as a central regulator of metabolic flux under different conditions. These findings provide targeted insights into the complex metabolic network governing rice grain chalkiness under WNT conditions. Integration of GSM and transcriptomics data, enhanced our understanding of the intricate relationship between environmental factors, metabolic processes, and grain quality and also offer markers that can be useful to develop rice with improved resilience.

Suboptimal grain filling in rice (Oryza sativa) inferior spikelets poses a constraint to maximizing the yield potential. The differential grain filling between superior and inferior spikelets has been primarily attributed to differences in endogenous phytohormone levels that determine grain sink capacity. In this study, we aimed to gain molecular insights into the role of two phytohormones, abscisic acid (S-ABA or ABA) and cytokinin (6-benzyladenine or BA) through exogenous applications on superior and inferior grains in rice. We found that ABA and a combination of ABA and BA (ABA+BA) applications increased the grain yield in field studies, primarily by improving the grain weight of both superior and inferior grains. Transcriptomic analysis of developing grains shows differences in the expression of core cell cycle genes between the superior and inferior grains at four days after fertilization between the control and phytohormone applications. ABA and ABA+BA applications induce DNA replication genes and cell cycle inhibitory genes in superior grains only, likely promoting endoreduplication for increased cell storage capacity. ABA and BA applications suppressed the expression of cytokinin signaling genes in superior grains but induced them in inferior grains emphasizing the key roles for cytokinins and ABA in superior and inferior grains, respectively. An early induction of several grain storage-related genes in inferior grains is associated with accelerated entry into the grain storage stage, thus limiting sink capacity and poor grain fill. Our results indicate that ABA alone promotes photosynthate remobilization into both superior and inferior grains while ABA + BA regulates grain filling via cell cycle-related transcriptomic changes. Overall, our study reveals an intrinsic difference in ABA+BA sensitivity between inferior and superior grains that is linked to regulation of cell cycle checkpoints and entry into endoreduplication in the endosperm.

Phenotyping of crop panicles is crucial for identifying traits that influence yield and grain quality; however, the complexity and scale of modern phenotyping data pose significant analytical challenges. Modern phenotyping data are inherently multidimensional, combining spatial, genotypic, treatment, and temporal variations, making interpretation increasingly challenging and motivating scalable visualization approaches. Effective visualization is, therefore, critical for detecting patterns and guiding biological discovery. Existing approaches often treat all dimensions equally or rely on dimensionality reduction, which may obscure key relationships. Limited screen space further constrains effective analysis. To address these challenges, we propose a multidimensional and multilevel graph (MMG) model, where nodes represent hierarchical components of the panicle (panicle, branch, seed) and features are analyzed both within and across levels. Each node also supports exploration across genotype, treatment, and time. Building on this model, we introduce PanicleVis, a scalable visual analytics system that integrates relationships across dimensions and levels while optimizing screen space to maintain clarity. We evaluate PanicleVis with real-world rice datasets collected under diverse conditions. Expert feedback confirms both the interpretability of MMG and the system's effectiveness in handling large-scale, multidimensional data. Its flexible design also supports feature extensions, establishing PanicleVis as a scalable framework for complex phenotyping studies.

It has been argued that metabolic content can be used as a selection marker to accelerate crop improvement because metabolic profiles in crops are often under genetic control. Evaluating the role of genetics in metabolic variation is a long-standing challenge. Rice, one of the world's most important staple crops, is known to be sensitive to recent increases in nighttime temperatures. Quantification of metabolic levels can help measure rice responses to high night temperature (HNT) stress. However, the extent of metabolic variation that can be explained by regression on whole-genome molecular markers remains to be evaluated. In the current study, we generated metabolic profiles for mature grains from a subset of rice diversity panel accessions grown under optimal and HNT conditions. Metabolite accumulation was low to moderately heritable, and genomic prediction accuracies of metabolite accumulation were within the expected upper limit set by their genomic heritability estimates. Genomic heritability estimates were slightly higher in the control group than in the HNT group. Genomic correlation estimates for the same metabolite accumulation between the control and HNT conditions indicated the presence of genotype-by-environment interactions. Reproducing kernel Hilbert spaces regression and image-based deep learning improved prediction accuracy, suggesting that some metabolite levels are under non-additive genetic control. Joint analysis of multiple metabolite accumulation simultaneously was effective in improving prediction accuracy by exploiting correlations among metabolites. The current study serves as an important first step in evaluating the cumulative effect of markers in influencing metabolic variation under control and HNT conditions.

Global nighttime temperatures are rising at twice the rate of daytime temperatures and pose a challenge for rice (Oryza sativa) production. High nighttime temperature (HNT) stress affects rice yield by reducing grain weight, size, and fertility. Although the genes associated with these yield parameters have been identified and characterized under normal temperatures, the genetic basis of grain weight regulation under HNT stress remains less explored. We examined the natural variation for rice single grain weight (SGW) under HNT stress imposed during grain development. A genome-wide association analysis identified several loci associated with grain weight under HNT stress. A locus, SGW1, specific to HNT conditions resolved to LONELY GUY-Like 1 (LOGL1), which encodes a putative cytokinin-activation enzyme. We demonstrated that LOGL1 contributes to allelic variation at SGW1. Accessions with lower LOGL1 transcript abundance had higher grain weight under HNT. This was supported by the higher grain weight of logl1-mutants relative to the wild type under HNT. Compared to logl1-mutants, LOGL1 over-expressers showed increased sensitivity to HNT. We showed that LOGL1 regulates the thiamin biosynthesis pathway, which is under circadian regulation, which in turn is likely perturbed by HNT stress. These findings provide a genetic source to enhance rice adaptation to warming night temperatures and improve our mechanistic understanding of HNT stress tolerance pathways.

Heat stress (HS) occurring during the grain-filling period has a detrimental effect on grain yield and quality in rice (Oryza sativa). The development of heat-resilient cultivars could partly solve this issue if tolerant alleles can be identified and incorporated into the germplasm. In this study, we posit that some of the phenotypic variations for heat resilience during grain development could be due to variations in gene expression among accessions. To test this, we characterized the HS response of 10 diverse rice accessions from three major sub-populations using physiological and transcriptome analyses. At a single-grain level, grain width and grain thickness emerged as the most heat-sensitive traits. During a transient HS, IND-3 was categorized as highly sensitive, while five accessions exhibited moderate heat sensitivity, and four accessions were tolerant. Only a core set of 29.4% of the differentially expressed genes was common to the three rice sub-populations. Heat-tolerant accession TEJ-5 uniquely triggered an unfolded protein response (UPR) under HS, as evident from the induction of OsbZIP50 and downstream UPR genes. OsbZIP58, a gene that positively regulates grain filling, was more highly induced by HS in IND-2 despite its moderate heat sensitivity. Collectively, our analysis suggests that both unique gene expression responses and variation in the level of responses for a given pathway distinguish diverse accessions. Only some of these responses are associated with single-grain phenotypes in a manner consistent with the known roles of these genes and pathways.

IntroductionSeed vigor is largely a product of sound seed development, maturation processes, genetics, and storage conditions. It is a crucial factor impacting plant growth and crop yield and is negatively affected by unfavorable environmental conditions, which can include drought and heat as well as cold wet conditions. The latter leads to slow germination and increased seedling susceptibility to pathogens. Prior research has shown that a class of plant growth regulators called substituted tertiary amines (STAs) can enhance seed germination, seedling growth, and crop productivity. However, inconsistent benefits have limited STA adoption on a commercial scaleMethodsWe developed a novel seed treatment protocol to evaluate the efficacy of 2-(N-methyl benzyl aminoethyl)-3-methyl butanoate (BMVE), which has shown promise as a crop seed treatment in field trials. Transcriptomic analysis of rice seedlings 24 h after BMVE treatment was done to identify the molecular basis for the improved seedling growth. The impact of BMVE on seed development was also evaluated by spraying rice panicles shortly after flower fertilization and subsequently monitoring the impact on seed traits.ResultsBMVE treatment of seeds 24 h after imbibition consistently improved wheat and rice seedling shoot and root growth in lab conditions. Treated wheat seedlings grown to maturity in a greenhouse also resulted in higher biomass than controls, though only under drought conditions. Treated seedlings had increased levels of transcripts involved in reactive oxygen species scavenging and auxin and gibberellic acid signaling. Conversely, several genes associated with increased reactive oxygen species/ROS load, abiotic stress responses, and germination hindering processes were reduced. BMVE spray increased both fresh and mature seed weights relative to the control for plants exposed to 96 h of heat stress. BMVE treatment during seed development also benefited germination and seedling growth in the next generation, under both ambient and heat stress conditions.DiscussionThe optimized experimental conditions we developed provide convincing evidence that BMVE does indeed have efficacy in plant growth enhancement. The results advance our understanding of how STAs work at the molecular level and provide insights for their practical application to improve crop growth.

BACKGROUND:Wheat grain development in the first few days after pollination determines the number of endosperm cells that influence grain yield potential and is susceptible to various environmental conditions, including high night temperatures (HNTs). Flag leaves and seed-associated bracts (glumes, awn, palea, and lemma) provide nutrients to the developing seed. However, the specific metabolic roles of these tissues are uncertain, especially their dynamics at different developmental stages and the time in a day. Tissue- and time-dependent metabolite profiling may hint at the metabolic roles of tissues and the mechanisms of how HNTs affect daytime metabolic status in early grain development.RESULTS:The metabolite profiles of flag leaf, bract, seed (embryo and endosperm), and entire spike were analyzed at 12:00 (day) and 23:00 (night) on 2, 4, and 6 days after fertilization under control and HNT conditions. The metabolite levels in flag leaves and bracts showed day/night oscillations, while their behaviors were distinct between the tissues. Some metabolites, such as sucrose, cellobiose, and succinic acid, showed contrasting oscillations in the two photosynthetic tissues. In contrast, seed metabolite levels differed due to the days after fertilization rather than the time in a day. The seed metabolite profile altered earlier in the HNT than in the control condition, likely associated with accelerated grain development caused by HNT. HNT also disrupted the day/night oscillation of sugar accumulation in flag leaves and bracts.CONCLUSIONS:These results highlight distinct metabolic roles of flag leaves and bracts during wheat early seed development. The seed metabolite levels are related to the developmental stages. The early metabolic events in the seeds and the disruption of the day/night metabolic cycle in photosynthetic tissues may partly explain the adverse effects of HNT on grain yield.

Abstract High night air temperature (HNT) stress negatively impacts both rice (Oryza sativa L) yield and grain quality and has been extensively investigated because of the significant yield loss observed (10%) for every increase in air temperature (1°C). Most of the rice HNT studies have been conducted under greenhouse conditions, with limited information on field‐level responses for the major rice sub‐populations. This is due to a lack of a field‐based phenotyping infrastructure that can accommodate a diverse set of accessions representing the wider germplasm and impose growth stage‐specific stress. In this study, we built six high‐tunnel greenhouses and screened 310 rice accessions from the Rice Diversity Panel 1 (RDP1) and 10 commercial hybrid cultivars in a replicated design. Each greenhouse had heating and a cyber–physical system that sensed ambient air temperature and automatically increased night air temperature to about 4°C relative to ambient temperature in the field for two cropping seasons. The system successfully imposed HNT stress of 4.0 and 3.94°C as recorded by Raspberry Pi sensors for 2 weeks in 2019 and 2020, respectively. HOBO sensors (Onset Computer Corporation) recorded a 2.9 and 2.07°C temperature differential of ambient air between control and heated greenhouses in 2019 and 2020, respectively. These greenhouses were able to withstand constant flooding, heavy rains, strong winds (140 mph), and thunderstorms. Selected US rice cultivars showed an average of 24% and 15% yield reduction under HNT during the 2019 and 2020 cropping seasons, respectively. Our study highlights the potential of this computer‐based infrastructure for accurate implementation of HNT or other abiotic stresses under field‐growing conditions.

A framework combining two powerful tools of hyperspectral imaging and deep learning for the processing and classification of hyperspectral images (HSI) of rice seeds is presented. A seed-based approach that trains a three-dimensional convolutional neural network (3D-CNN) using the full seed spectral hypercube for classifying the seed images from high day and high night temperatures, both including a control group, is developed. A pixel-based seed classification approach is implemented using a deep neural network (DNN). The seed and pixel-based deep learning architectures are validated and tested using hyperspectral images from five different rice seed treatments with six different high temperature exposure durations during day, night, and both day and night. A stand-alone application with Graphical User Interfaces (GUI) for calibrating, preprocessing, and classification of hyperspectral rice seed images is presented. The software application can be used for training two deep learning architectures for the classification of any type of hyperspectral seed images. The average overall classification accuracy of 91.33% and 89.50% is obtained for seed-based classification using 3D-CNN for five different treatments at each exposure duration and six different high temperature exposure durations for each treatment, respectively. The DNN gives an average accuracy of 94.83% and 91% for five different treatments at each exposure duration and six different high temperature exposure durations for each treatment, respectively. The accuracies obtained are higher than those presented in the literature for hyperspectral rice seed image classification. The HSI analysis presented here is on the Kitaake cultivar, which can be extended to study the temperature tolerance of other rice cultivars.

Graph convolutional neural network architectures combine feature extraction and convolutional layers for hyperspectral image classification. An adaptive neighborhood aggregation method based on statistical variance integrating the spatial information along with the spectral signature of the pixels is proposed for improving graph convolutional network classification of hyperspectral images. The spatial-spectral information is integrated into the adjacency matrix and processed by a single-layer graph convolutional network. The algorithm employs an adaptive neighborhood selection criteria conditioned by the class it belongs to. Compared to fixed window-based feature extraction, this method proves effective in capturing the spectral and spatial features with variable pixel neighborhood sizes. The experimental results from the Indian Pines, Houston University, and Botswana Hyperion hyperspectral image datasets show that the proposed AN-GCN can significantly improve classification accuracy. For example, the overall accuracy for Houston University data increases from 81.71% (MiniGCN) to 97.88% (AN-GCN). Furthermore, the AN-GCN can classify hyperspectral images of rice seeds exposed to high day and night temperatures, proving its efficacy in discriminating the seeds under increased ambient temperature treatments.

Water scarcity is the primary environmental constraint affecting wheat growth and production and is increasingly exacerbated due to climatic fluctuation, which jeopardizes future food security. Most breeding efforts to improve wheat yields under drought have focused on above-ground traits. Root traits are closely associated with various drought adaptability mechanisms, but the genetic variation underlying these traits remains untapped, even though it holds tremendous potential for improving crop resilience. Here, we examined this potential by re-introducing ancestral alleles from wild emmer wheat (Triticum turgidum ssp. dicoccoides) and studied their impact on root architecture diversity under terminal drought stress. We applied an active sensing electrical resistivity tomography approach to compare a wild emmer introgression line (IL20) and its drought-sensitive recurrent parent (Svevo) under field conditions. IL20 exhibited greater root elongation under drought, which resulted in higher root water uptake from deeper soil layers. This advantage initiated at the pseudo-stem stage and increased during the transition to the reproductive stage. The increased water uptake promoted higher gas exchange rates and enhanced grain yield under drought. Overall, we show that this presumably 'lost' drought-induced mechanism of deeper rooting profile can serve as a breeding target to improve wheat productiveness under changing climate.

Leaf numbers are vital in estimating the yield of crops. Traditional manual leaf-counting is tedious, costly, and an enormous job. Recent convolutional neural network-based approaches achieve promising results for rosette plants. However, there is a lack of effective solutions to tackle leaf counting for monocot plants, such as sorghum and maize. The existing approaches often require substantial training datasets and annotations, thus incurring significant overheads for labeling. Moreover, these approaches can easily fail when leaf structures are occluded in images. To address these issues, we present a new deep neural network-based method that does not require any effort to label leaf structures explicitly and achieves superior performance even with severe leaf occlusions in images. Our method extracts leaf skeletons to gain more topological information and applies augmentation to enhance structural variety in the original images. Then, we feed the combination of original images, derived skeletons, and augmentations into a regression model, transferred from Inception-Resnet-V2, for leaf-counting. We find that leaf tips are important in our regression model through an input modification method and a Grad-CAM method. The superiority of the proposed method is validated via comparison with the existing approaches conducted on a similar dataset. The results show that our method does not only improve the accuracy of leaf-counting, with overlaps and occlusions, but also lower the training cost, with fewer annotations compared to the previous state-of-the-art approaches.The robustness of the proposed method against the noise effect is also verified by removing the environmental noises during the image preprocessing and reducing the effect of the noises introduced by skeletonization, with satisfactory outcomes.