| Abstract Detail
The Potential of Machine Learning for Plant Biology SINGH, ARTI [1]. Human in the loop machine learning applications in high throughput plant stress phenotyping. Machine Learning (ML) approaches have emerged as a versatile tool to assimilate large amounts of heterogeneous data and provide reliable predictions of complex and uncertain phenomena. ML tools can work with big data obtained from complex and integrated phenotyping platforms like unmanned aerial vehicles, rovers and smartphones to solve the identification, classification, quantification, and prediction paradigm of plant stress phenotyping to generate insights that were previously not possible. The next step is to implement computer aided stress phenotyping for actionable outcomes in digital agriculture. We demonstrate a deep convolution neural network framework that can identify, classify and quantify diverse set of eight foliar stresses including fungal, bacterial, nutrient deficiency and herbicide injury in soybean [Glycine max (L.) Merr.]. To satisfy the practitioner's confidence in artificial intelligence generated ratings, we present explainability of the model using the top-K high-resolution feature maps that isolate the visual symptoms. This unsupervised identification of visual symptoms provides a quantitative measure of stress severity, allowing for identification (type of foliar stress), classification (low, medium, or high stress), and quantification (stress severity) providing automation. We successfully report transfer learning as the model is able to be deployed in non-soybean species. Overall these approaches are amenable to high throughput and fidelity aerial phenotyping platform as well as mobile platforms facilitating research and crop production solutions and is a current focus of our work. Log in to add this item to your schedule
1 - Iowa State University, 2104 Agronomy Hall , 716 Farm House Lane , Ames, IA , 50011, USA
Keywords: phenotyping deep learning.
Presentation Type: Symposium Presentation Session: SYM1, The Potential of Machine Learning for Plant Biology Location: Tucson F/Starr Pass Date: Monday, July 29th, 2019 Time: 11:15 AM Number: SYM1007 Abstract ID:973 Candidate for Awards:Margaret Menzel Award,Edgar T. Wherry award |