Train Robots dataset is designed to develop systems capable of understanding the verbal spatial commands described in a natural way (Dukes 2013). Non-expert users were asked to annotate appropriate commands to 1000 pairs of different scenes: 7752 commands were collected using Amazon Mechanical Turk. We also translated all the commands from English to Arabic, particular care was taken on not to alter any command or change any mistakes in any of them.

Data and Resources

Additional Info

Field Value
Author Kais Dukes, Muhannad Alomari
Maintainer Muhannad Alomari