Crowdsourcing image analysis for plant phenomics to generate ground truth data for machine learning

Authors: ZHOU, NAIHUI - Iowa State University; SIEGEL, ZACHARY - Iowa State University; ZARECOR, SCOTT - Iowa State University; LEE, NIGEL - Iowa State University; CAMPBELL, DARWIN - Iowa State University; Andorf, Carson; NETTLETON, DAN - Iowa State University; LAWRENCE-DILL, CAROLYN - Iowa State University; GANAPATHYSUBRAMANIAN, BASKAR - Iowa State University; KELLY, JONATHAN - Iowa State University; FRIEDBERG, IDDO - Iowa State University

Submitted to: PLoS Computational Biology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/23/2018
Publication Date: 7/30/2018
Citation: Zhou, N., Siegel, Z.D., Zarecor, S., Lee, N., Campbell, D.A., Andorf, C.M., Nettleton, D., Lawrence-Dill, C.J., Ganapathysubramanian, B., Kelly, J.W., Friedberg, I. 2018. Crowdsourcing image analysis for plant phenomics to generate ground truth data for machine learning. PLoS Computational Biology. 14(7):e1006337. https://doi.org/10.1371/journal.pcbi.1006337.
DOI: https://doi.org/10.1371/journal.pcbi.1006337

Interpretive Summary: A major challenge in machine learning is the lack of affordable high-quality ground-truth data sets. These data sets are necessary to build models for predictions. Generating these data sets are especially challenging in life sciences where experimental determination is both time consuming and expensive. Here we explore the use of public (Amazon MTurk) and private (students) crowdsourcing techniques to generate a large number of good quality training data involving the identification of corn tassels from images taken in a field setting. We conclude that properly managed crowdsourcing can be used to establish large volumes of viable ground truth data at a low cost and high quality, especially in the context of high throughput image analysis.

Technical Abstract: The accuracy of machine learning tasks critically depends on high quality ground truth data. Therefore, in many cases, producing good ground truth data typically involves trained professionals; however, this can be costly in time, effort, and money. Here we explore the use of crowdsourcing to generate a large number of training data of good quality. We explore an image analysis task involving the segmentation of corn tassels from images taken in a field setting. We investigate the accuracy, speed and other quality metrics when this task is performed by students for academic credit, Amazon MTurk workers, and Master Amazon MTurk workers. We conclude that the Amazon MTurk and Master Mturk workers perform significantly better than the for-credit students, but with no significant difference between the two MTurk worker types. Furthermore, the quality of the segmentation produced by Amazon MTurk workers rivals that of an expert worker. We provide best practices to assess the quality of ground truth data, and to compare data quality produced by different sources. We conclude that properly managed crowdsourcing can be used to establish large volumes of viable ground truth data at a low cost and high quality, especially in the context of high throughput plant phenotyping. We also provide several metrics for assessing the quality of the generated datasets.