/Perceptions_MTurk

Colect perceptions using Amazon mechanical turk

Primary LanguageJupyter NotebookMIT LicenseMIT

Perceptions_MTurk

Collect perceptions of the urban built environment using Amazon mechanical turk

Author: Andrew Larkin
Principal Investigator: Perry Hystad
Summary: Create standardized methodology for capturing perceptions of street view locations using Google Street View and Amazon Mechanical Turk. Novel components of standardized methodology include:

  • adjust for differences between road and compass heading while standardizing viewing angles
  • creating training image sets that maximize distribution of built environment features
  • select image comparison pairs with dyanmically updated trueskill scores
  • adjust for personal bias using QA/QC questions and intravote variance for each participant

Repository Files

  • dataFiles - the training dataset and documentation files
  • statisticalAnalysis - analyze developed methods and output products
  • surveyBackend - server used to distribute sampled image comparisons and collect participant respones
  • surveyClient - React-based client to run survey in web browser
  • lib - library of custom classes
  • imageSelection - methods to sample and select GSV images for the training dataset

Additional Resources

  1. Spatial Health Lab - Oregon State University
  2. Amazon Mechanical Turk - used to recruit participants to the study
  3. Google Street View - standardized database of street view imagery
  4. Todo: enter link to Ajay's github repo