Speaker: Alex Sorokin
 
Affiliation: UIUC (CS)

Title:

Data annotation with Amazon Mechanical Turk

 
Abstract:
Abstract: Annotated data is an important cornerstone for developing theories and algorithms in 
Computer Vision. In this talk I will present how to build large annotated datasets using Amazon 
Mechanical Turk. Annotated data can then be obtained in large quantities quite cheaply. I will 
discuss multiple use cases and my own experience. Setting up a successful annotation project 
requires a number of questions answered. I will discuss how to control quality, how to set the 
pay for the tasks and what properties of the tasks make them succeed. 
 
 
back