No More Manual Monitoring of Plant Emergence. New AI System Saves Time and Provides Data

Scientists at CATRIN, Palacký University, in collaboration with colleagues from the Czech University of Life Sciences Prague, have developed a practical tool for researchers, plant breeders, and farmers. The SPROUT system (AI-based Seedling PRedictiOn and trait extraction Using RGB Time-series), which relies on AI-assisted analysis of RGB images over time, automatically detects the time of crop emergence and derives biologically significant parameters from it. The authors presented this innovation in a study published in the prestigious Computers and Electronics in Agriculture journal.

“Early crop development significantly influences plant growth and final yield; therefore, seedling emergence is an important parameter in monitoring crop condition, plant breeding, and studying plant responses to stress. However, its assessment is still often performed manually and typically records only the final percentage of emerged plants, resulting in the loss of valuable information about the emergence process. In contrast, SPROUT, which we developed, uses artificial intelligence and is capable of automatically tracking, from standard RGB photographs, when and how uniformly the plants are emerging,” said Nuria De Diego, research leader at CATRIN.

The researchers developed and tested the system using barley and wheat data captured by various RGB cameras under controlled conditions in growth chambers. “During model development and retraining, the most successful model—a Temporal Convolutional Network (TCN)—achieved 90 percent accuracy in determining the time of emergence, which enabled a highly accurate reconstruction of emergence curves. When tested on a completely independent dataset without further retraining, the model was still able to capture the basic dynamics of sprouting, albeit with lower accuracy. This shows that SPROUT is best used as a modular system that can be easily retrained or fine-tuned for new crops, cameras, or experimental conditions,” explained the article’s first author, Pavel Klimeš.

The researchers also conducted a case study focusing on cadmium stress in two different wheat genotypes. It showed that different genotypes exhibit distinct germination patterns in terms of speed and synchrony, which is not random but reflects their biological characteristics and ability to respond to the environment.

SPROUT is currently a research software tool ready for collaboration with the applied sector. Thanks to its modular design, it can be further customized and trained for various crops, cameras, or experimental conditions. The system has particular potential for screening genotypes in plant breeding and seed production, as well as for testing the effects of biostimulants, fertilizers, and agrochemicals on early plant development.

“The next step will be to expand and validate the system across a broader range of crops, substrates, and imaging systems. The goal is to develop SPROUT into an easy-to-use tool that would make automated analysis of plant germination accessible to breeders, seed companies, manufacturers of agricultural products, and research institutions without requiring in-depth knowledge of artificial intelligence,” concluded Lukáš Spíchal, one of the study’s authors and head of CATRIN-BIO.