Challenge 1: Machine Learning approaches to High Throughput Phenotyping



Stanton Martin and Ambarish Nag

BioSciences Division at Oak Ridge National Laboratory



Plant phenotyping refers to a quantitative description of the plant’s anatomical, ontogenetical, physiological and biochemical properties. Today, rapid developments are taking place in the field of non-destructive, image-analysis -based phenotyping that allow for a characterization of plant traits in high-throughput. During the last decade, ‘the field of image-based phenotyping has broadened its focus from the initial characterization of single-plant traits in controlled conditions towards ‘real-life’ applications of robust field techniques in plant plots and canopies1.  A grand challenge in the field of plant phenotyping are the extraction of biologically relevant features from large datasets generated by robotic, field based instrumentation.  Machine learning, as well as traditional segmentation approaches have been used for this task.  Software packages such as “Greenotyper” utilize deep learning neural networks combined with traditional object detection methods such as thresholding to segment and process large amounts of imagery.2


This dataset consists of labeled images from Populus Trichocarpa genotypes cultivated under both drought and control conditions at a common garden located in California.  The treatment, block, row, position, and genotype are indicated on the tag. The images were collected using cell phones connected to high precision GPS instrumentation at a spatial resolution of 10 centimeters or better. The dataset can be accessed at:

 The challenges for this dataset are:

  1. Is it possible to use optical character recognition (OCR) or machine learning techniques to “Read” the label on each tag and generate a spreadsheet contains the treatment, block, row, position, and genotype? Doing this would dramatically simplify data collection, as this information is usually collected manually.
  2. Can machine learning differentiate and classify different leaf morphologies among genotypes by classifying leaf shape or color characteristics?
  3. Can a predictive model be built using leaf morphology classifications that may indicate that a particular genotype was cultivated in a “drought” or “control” condition?
  4. GPS and other camera information are encoded in exif tags. Can this data be used to determine characteristics such as leaf size? Can other data, such as soil maps, weather, etc. be used to find correlations among phenotypes?


  1. Walter, A., Liebisch, F. & Hund, A. Plant phenotyping: from bean weighing to image analysis.Plant Methods 11, 14 (2015).
  1. Tausen, Marni, Clausen, Marc, Moeskjaer,Sara,Shihavuddin,ASM, Dahl, Anders Bjorholm, Janss, Luc, Anderson, Stig Uggerhoj. Greenotyper: Image-Based Plant Phenotyping Using Distributed Computing and Deep Learning. Frontiers in Plant Science. 11 (2020.