the-midnight-paper

Tohme: Detecting Curb Ramps in Google Street View using Crowdsourcing, Computer Vision, and Machine Learning

Authors: Kotaro Hara, Jin Sun, Robert Moore, David Jacobs, Jon E. Froehlich

Conference: UIST ‘14, October 05 - 08 2014, Honolulu, HI, USA

Keywords: Crowdsourcing accessibility; computer vision; Google Street View; Amazon Mechanical Turk

Strength

One of the biggest strength of TohMe is Dynamic Workflow Allocation via svControl. This machine-learning module not only schedules work via Performance Prediction, but also assists in predicting CV performance and assignment of work to either a manual labor pipeline (svLabel) or an automated pipeline with human verification.

Weakness

Crowd-based solution to collect information about the accessibility of the built environment incurs additional costs. While authors have addressed this point in the paper, their argument on approaching families of patients, is not completely convincing.

Future Work

It would be interesting to gamify the interface for human verification and labeling of curb ramps and see how such gamification techniques effect worker’s accuracy.