Abstract
Laboratory experiments for human behavioral studies lack scalability or generality, making big cognitive studies non-trivial especially with complex natural scenes or less controlled environment. To tackle this issue, we propose a web-based behavioral experiment platform for crowdsourcing large-scale behavioral data. The main target of the platform is focused on the area of visual attention. It provides an inexpensive and easy-to-use method for scalable data collection relating to image perception and visual attention. Core features include a mouse-tracking experiment to simulate eye tracking with mouse or finger movements, an image annotation experiment to characterize complex stimuli, and a general survey. Experimenters set up studies on this platform using customized templates of these experiments. The platform also allows experimenters to publish their experiments on Amazon Mechanical Turk with simple configurations. Experimental results are available for download and visualization at any time.
Meeting abstract presented at VSS 2017