Explainable AI in Plant Phenotyping: A Critical Component of Food Security

Researchers at the University of Saskatchewan have underscored the importance of explainable artificial intelligence (XAI) in plant phenotyping, a crucial step towards ensuring global food security. As the world's population continues to grow, and climate change poses an increasingly significant threat to crop yields, breeders and growers need tools to develop and manage crop cultivars that can withstand extreme weather conditions. Plant phenotyping, which involves the measurement of a plant's structural and functional characteristics, can inform and improve breeding and management decisions. However, current deep learning methods used for image-based phenotyping remain difficult to interpret, making them a "black box." XAI has the potential to open this black box, providing plant scientists with trustworthy and interpretable data.

Key Takeaways:

• The increasing human population and climate change pose a significant threat to global food security, emphasizing the need for innovative solutions to improve crop resilience.
• Plant phenotyping, the measurement of a plant's structural and functional characteristics, is a critical component of breeding and management decisions for growers.
• Current deep learning methods used for image-based phenotyping are difficult to interpret, limiting their trustworthiness and efficacy.
• Explainable AI (XAI) has the potential to enhance the trustworthiness and interpretability of image-based phenotypic information, making it a promising approach for plant phenotyping research.
• Researchers at the University of Saskatchewan have conducted a comprehensive review of XAI studies in plant shoot phenotyping and related domains to highlight the benefits of XAI and encourage its integration into future studies.
• The reviewed XAI studies demonstrate the potential of XAI to elucidate the representations within deep learning models, enabling researchers to explain model decisions, relate detected features to underlying plant physiology, and enhance the trustworthiness of image-based phenotypic information.

Statistics:

• 6 (year of publication) - identifier for the research paper published in Frontiers in Artificial Intelligence.
• October 2, 2023 (date of publication) - the date on which the news report was published.
• 1010 (article number) - the article number of the news report.
• 6 (number of authors) - the total number of authors involved in the research.
• 3 (number of research institutions) - the number of research institutions involved in the study.

Sources:

• NewsRx. University of Saskatchewan Researchers Detail New Studies and Findings in the Area of Artificial Intelligence (Explainable deep learning in plant phenotyping). Robotics & Machine Learning. October 2, 2023; p 1010.
• Mostafa, S., Mondal, D., Panjvani, K., Kochian, L., Stavness, I. (2023). Explainable deep learning in plant phenotyping. Frontiers in Artificial Intelligence, 6.