APPLIED CROWDSOURCING: ZEBRAFISH PHENOTYPING AND BRUSHSTROKE EXTRACTION
Open Access
- Author:
- Schleicher, John Cassidy
- Area of Honors:
- Computer Science
- Degree:
- Bachelor of Science
- Document Type:
- Thesis
- Thesis Supervisors:
- James Z Wang, Thesis Supervisor
James Z Wang, Thesis Supervisor
John Hanna, Thesis Honors Advisor - Keywords:
- crowdsourcing
phenotype
phenotyping
human-based computation
image analysis
annotation - Abstract:
- Currently, no tool exists which can automatically annotate biological images with a high level of precision and accuracy. Until a tool is developed that can automate this process, it is necessary to use alternative methods to complete this task, such as manually annotating biological images. The goal of this thesis is to develop an online crowdsourcing tool to annotate biological images, specifically images of zebrafish. The implementation of this tool, named CABERNET, will enable the creation of a knowledgebase for users who need this type of information. The main challenges that were overcome in the development of CABERNET were those of security, usability, and quality assurance. Preliminary testing of CABERNET has shown that it is 14% faster on average than using spreadsheets to record annotations. These preliminary results of CABERNET’s usage demonstrate that a crowdsourcing approach has meaningful applications in the realm of image annotation. Another project titled the van Gogh project is also covered in this thesis. In it, human annotation is used to extract brushstrokes from images of the paintings of van Gogh and his contemporaries. The annotation and analytical techniques used created a dataset which is used for comparison with an automated tool. Both the van Gogh and CABERNET projects are capable of generating valuable datasets which can be used in further research.