RESUMO

Common beans (CB), a vital source for high protein content, plays a crucial role in ensuring both nutrition and economic stability in diverse communities, particularly in Africa and Latin America. However, CB cultivation poses a significant threat to diseases that can drastically reduce yield and quality. Detecting these diseases solely based on visual symptoms is challenging, due to the variability across different pathogens and similar symptoms caused by distinct pathogens, further complicating the detection process. Traditional methods relying solely on farmers' ability to detect diseases is inadequate, and while engaging expert pathologists and advanced laboratories is necessary, it can also be resource intensive. To address this challenge, we present a AI-driven system for rapid and cost-effective CB disease detection, leveraging state-of-the-art deep learning and object detection technologies. We utilized an extensive image dataset collected from disease hotspots in Africa and Colombia, focusing on five major diseases: Angular Leaf Spot (ALS), Common Bacterial Blight (CBB), Common Bean Mosaic Virus (CBMV), Bean Rust, and Anthracnose, covering both leaf and pod samples in real-field settings. However, pod images are only available for Angular Leaf Spot disease. The study employed data augmentation techniques and annotation at both whole and micro levels for comprehensive analysis. To train the model, we utilized three advanced YOLO architectures: YOLOv7, YOLOv8, and YOLO-NAS. Particularly for whole leaf annotations, the YOLO-NAS model achieves the highest mAP value of up to 97.9% and a recall of 98.8%, indicating superior detection accuracy. In contrast, for whole pod disease detection, YOLOv7 and YOLOv8 outperformed YOLO-NAS, with mAP values exceeding 95% and 93% recall. However, micro annotation consistently yields lower performance than whole annotation across all disease classes and plant parts, as examined by all YOLO models, highlighting an unexpected discrepancy in detection accuracy. Furthermore, we successfully deployed YOLO-NAS annotation models into an Android app, validating their effectiveness on unseen data from disease hotspots with high classification accuracy (90%). This accomplishment showcases the integration of deep learning into our production pipeline, a process known as DLOps. This innovative approach significantly reduces diagnosis time, enabling farmers to take prompt management interventions. The potential benefits extend beyond rapid diagnosis serving as an early warning system to enhance common bean productivity and quality.

Assuntos

Aprendizado Profundo , Phaseolus , Doenças das Plantas , Phaseolus/virologia , Phaseolus/microbiologia , Doenças das Plantas/virologia , Doenças das Plantas/microbiologia , Agricultura/métodos , Folhas de Planta/virologia , Folhas de Planta/microbiologia , África , Colômbia

RESUMO

Common bean (Phaseolus vulgaris L.) is the second most important source of dietary protein and the third most important source of calories in Africa, especially for the poor. In East Africa, drought is an important constraint to bean production. Therefore, breeding programs in East Africa have been trying to develop drought resistant varieties of common bean. To do this, breeders need information about seasonal drought stress patterns including their onset, intensity, and duration in the target area of the breeding program, so that they can mimic this pattern during field trials. Using the Decision Support for Agrotechnology Transfer (DSSAT) v4.7 model together with historical and future (Coupled Model Inter-comparison Project 6, CMIP6) climate data, this study categorized Ethiopia, Tanzania, and Uganda into different target population of environments (TPEs) based on historical and future seasonal drought stress patterns. We find that stress-free conditions generally dominate across the three countries under historical conditions (50-80% frequency). These conditions are projected to increase in frequency in Ethiopia by 2-10% but the converse is true for Tanzania (2-8% reduction) and Uganda (17-20% reduction) by 2050 depending on the Shared Socioeconomic Pathway (SSP). Accordingly, by 2050, terminal drought stresses of various intensities (moderate, severe, extreme) are prevalent in 34% of Uganda, around a quarter of Ethiopia, and 40% of the bean growing environments in Tanzania. The TPEs identified in each country serve as a basis for prioritizing breeding activities in national programs. However, to optimize resource use in international breeding programs to develop genotypes that are resilient to future projected stress patterns, we argue that common bean breeding programs should focus primarily on identifying genotypes with tolerance to severe terminal drought, with co-benefits in relation to adaptation to moderate and extreme terminal drought. Little to no emphasis on heat stress is warranted by 2050s.

RESUMO

Root rot in common bean is a disease that causes serious damage to grain production, particularly in the upland areas of Eastern and Central Africa where significant losses occur in susceptible bean varieties. Pythium spp. and Fusarium spp. are among the soil pathogens causing the disease. In this study, a panel of 228 lines, named RR for root rot disease, was developed and evaluated in the greenhouse for Pythium myriotylum and in a root rot naturally infected field trial for plant vigor, number of plants germinated, and seed weight. The results showed positive and significant correlations between greenhouse and field evaluations, as well as high heritability (0.71-0.94) of evaluated traits. In GWAS analysis no consistent significant marker trait associations for root rot disease traits were observed, indicating the absence of major resistance genes. However, genomic prediction accuracy was found to be high for Pythium, plant vigor and related traits. In addition, good predictions of field phenotypes were obtained using the greenhouse derived data as a training population and vice versa. Genomic predictions were evaluated across and within further published data sets on root rots in other panels. Pythium and Fusarium evaluations carried out in Uganda on the Andean Diversity Panel showed good predictive ability for the root rot response in the RR panel. Genomic prediction is shown to be a promising method to estimate tolerance to Pythium, Fusarium and root rot related traits, indicating a quantitative resistance mechanism. Quantitative analyses could be applied to other disease-related traits to capture more genetic diversity with genetic models.

RESUMO

Common bean (Phaseolus vulgaris L.) is an important staple crop for smallholder farmers, particularly in Eastern and Southern Africa. To support common bean breeding and seed dissemination, a high throughput SNP genotyping platform with 1500 established SNP assays has been developed at a genotyping service provider which allows breeders without their own genotyping infrastructure to outsource such service. A set of 708 genotypes mainly composed of germplasm from African breeders and CIAT breeding program were assembled and genotyped with over 800 SNPs. Diversity analysis revealed that both Mesoamerican and Andean gene pools are in use, with an emphasis on large seeded Andean genotypes, which represents the known regional preferences. The analysis of genetic similarities among germplasm entries revealed duplicated lines with different names as well as distinct SNP patterns in identically named samples. Overall, a worrying number of inconsistencies was identified in this data set of very diverse origins. This exemplifies the necessity to develop and use a cost-effective fingerprinting platform to ensure germplasm purity for research, sharing and seed dissemination. The genetic data also allows to visualize introgressions, to identify heterozygous regions to evaluate hybridization success and to employ marker-assisted selection. This study presents a new resource for the common bean community, a SNP genotyping platform, a large SNP data set and a number of applications on how to utilize this information to improve the efficiency and quality of seed handling activities, breeding, and seed dissemination through molecular tools.