Sketch-and-test: picture-centered research with p5.js assisted crowdsourcing (2004.08198v3)

Abstract: Relating human judgements to pictures is central to a wide variety of scientific disciplines. Pictures are used to evoke and study faculties of the human mind, while human input is used to label, understand and model pictorial representations. Human input is often collected through online crowdsourcing experiments. This paper discusses the usage of crowdsourcing in two major branches of picture-centered research, human and computer vision, and identifies novel directions such as art history and design. We demonstrate that a wide variety of experiments can be conducted by using p5.js, a library originally intended to facilitate visual creation. We report five complementary experimental paradigms to illustrated the accessibility and versatility of p5.js: Change blindness, BubbleView, 3D shape perception, Composition, and Perspective reconstruction. Results reveal that literature findings can be reproduced and novel insights can easily be achieved with the p5.js library. The creative freedom of p5.js combined with low threshold access to crowdsourcing seems like a powerful combination for all picture-centred research areas: perception, design, art history, communication, and beyond.