Graduate Thesis Or Dissertation
 

Deep Segmentation in Plant Phenomics

Public Deposited

Downloadable Content

Download PDF
https://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/3j3339077

Descriptions

Attribute NameValues
Creator
Abstract
  • The abilities of plant biologists to characterize the genetic basis of physiological traits are limited by their abilities to obtain quantitative data representing precise details of trait variation and mainly to collect this data on a high-throughput scale at low cost. Deep learning-based methods have demonstrated unprecedented potential to automate plant phenotyping by replacing labor-intensive manual mode of measurements. Although these methods have yielded advances in phenotyping studies, their efficiency commonly relies on collecting large training sets that can be time-consuming to generate pixel-level annotations for studying plant traits and tissue stress response. Intelligent algorithms have been proposed that can enhance the productivity of these labeling tasks and reduce human efforts. However, most of these algorithms require careful labeling of the boundaries of the object of interest, which requires a significant amount of user effort. We propose an interactive object segmentation algorithm, Semantic-Guided Interactive object Segmentation (SGIS), which utilizes a semantic prior map and makes image annotation task much simpler. Our quantitative evaluation on Pascal VOC and Leaf Segmentation Challenge (LSC) shows that our proposed SGIS model requires fewer user inputs than the state-of-art models in interactive object segmentation. In addition, the results of a user study performed with the help of an expert biologist highlight that using semantic information is helpful for the interactive image segmentation task and can reduce the amount of time required to annotate a dataset. The plant biologists further performed a GWAS study with data labeled using our annotation GUI, Intelligent DEep Annotator for Segmentation (IDEAS), which revealed results in sync with the existing models of plant regeneration.
License
Resource Type
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Rights Statement
Publisher
Peer Reviewed
Language
Embargo reason
  • Pending Publication
Embargo date range
  • 2021-06-02 to 2022-07-02

Relationships

Parents:

This work has no parents.

In Collection:

Items

Thumbnail Title Date Uploaded Visibility Actions
file details KaurDamanpreet2021.pdf 2021-06-01 Public Download