Rice production is central to global food security, and flowering is a critical transition that influences pollination, seed set, yield and grain quality. Within a single day, floral events such as flower opening, anther dehiscence and stigma receptivity must be tightly coordinated, especially for hybrid breeding. Yet these traits are genetically complex and highly sensitive to temperature, radiation and other environmental fluctuations. Conventional manual or low-frequency observations struggle to capture rapid, often occluded flowering dynamics across hundreds of accessions, limiting genetic dissection and breeding applications. Given these challenges, deeper research is needed into automated, ultra-high-frequency phenotyping of diurnal flowering dynamics in rice.
In a study published (DOI: 10.1016/j.plaphe.2026.100215) in Plant Phenomics in 2026, researchers from Nanjing Agricultural University, the Shanghai Agrobiological Gene Center, the National Institute of Agricultural Botany (NIAB), the Chinese Academy of Agricultural Sciences and the Chinese Academy of Sciences introduced BloomSight. The platform combines time-lapse red-green-blue (RGB) imaging, deep-learning segmentation and time-series analysis to measure rice panicle and anther dynamics, identify flowering patterns and support genetic mapping.
BloomSight used a digital single-lens reflex (DSLR) camera on rotating turntables to capture images every few seconds from 07:00 to 14:00, collecting more than 530,000 pot-level images across 172 accessions. From these, the team established the Open Rice Flowering Training (ORFT) dataset, annotating 39,608 plant-level images, 5,009 panicles and over 350,000 anthers. A two-stage deep-learning pipeline—YOLACT-Panicle for panicle segmentation and UNet-Anther for anther detection—produced minute-level measures of anther area and number. Generalized additive model (GAM) curves, Kneedle detection and self-organizing map (SOM) clustering then defined initial flowering time (Tml), quickest flowering time (TQck), peak flowering time (TPeak), the rapid flowering phase (PRpd) and flowering density traits. The team classified accessions into Slow, Moderate and Fast patterns, with distinct flowering windows between japonica and indica. Time-series genome-wide association study (GWAS) using more than 1 million single nucleotide polymorphisms (SNPs) identified signals near known genes such as EMF1, OsMYB8, PME42 and EMT3, plus repeatedly detected unknown loci. Manually scored Tml and TPeak yielded no significant loci, whereas computationally derived dynamic traits did.
The authors said BloomSight turns a fast, easily missed biological process into a measurable minute-by-minute record. They said the framework can help breeders compare flowering windows, synchronize parental lines and design crosses with greater confidence. They added that combining high-frequency imaging, deep learning and genetics offers a practical way to study dynamic traits in rice and other cereals, while reducing reliance on subjective manual scoring. They said the approach is designed to be scalable and adaptable for broader crop research.
The findings could help rice breeders synchronize pollen release with stigma receptivity, improve hybrid seed production and avoid heat stress during flowering. Slow-pattern varieties may suit temperate climates, Moderate-pattern varieties may offer a useful flowering window for hybrid breeding, and Fast-pattern varieties may reduce heat-stress risk by narrowing flowering. The team said BloomSight could also support marker-assisted selection, gene editing and pre-breeding decisions. With further development, smartphone-based detection, edge computing and drone fleets could extend the framework to wheat, maize and other cereal crops. The open datasets and code are intended to help the broader plant research community study dynamic flowering traits.
###
References
DOI
10.1016/j.plaphe.2026.100215
Original Source URL
https://doi.org/10.1016/j.plaphe.2026.100215
Funding information
National Natural Science Foundation of China (U24A20399; 32070400), UK Research and Innovation (UKRI) Biotechnology and Biological Sciences Research Council (BBSRC) grants (BB/Y513969/1; BB/Y514081/1), Novo Nordisk Foundation Wheat Alliance programme (NNF24SA0096909), and Chinese Government Overseas Scholarship Council (202406850053).
About Plant Phenomics
Plant Phenomics publishes breakthrough research that integrates genomics, genetics, physiology, molecular biology, bioinformatics, statistics, mathematics and computer science to advance plant phenotyping—from the cellular level to entire populations. The journal tackles key scientific challenges in phenomics by showcasing cutting‑edge technologies for data acquisition, management, interpretation and modelling. Its scope spans practical applications across all plant sciences, aiming to translate novel phenotyping tools into real‑world solutions for breeding, crop management and fundamental plant biology. By fostering interdisciplinary innovation, Plant Phenomics serves as a hub for researchers developing high‑throughput, accurate and scalable methods to decipher complex plant traits, ultimately accelerating progress in agriculture and plant science.