Optimizing Character Animations using Online Crowdsourcing

Abstract

This paper presents a novel approach for exploring diverse and expressive motions that are physically correct and interactive. The approach combining user participation in with the animation development process using crowdsourcing to remove the need for data-driven libraries while address aesthetic limitations. A core challenge for character animation solutions that do not use pre-recorded data is they are constrained to specific actions or appear unnatural and out of place (compared to real-life movements). Character movements are very subjective to human perception (easily identify underlying unnatural or strange patterns with simple actions, such as walking or climbing). We present an approach that leverage's crowdsourcing to reduce these uncanny artifacts within generated character animations. Crowdsourcing animations is an uncommon practice due to the complexities of having multiple people working in parallel on a single animation. A web-based solution for analysis and animation is presented in this paper. It allows users to optimize and evaluate complicated character animation mechanism conveniently on-line. The context of this paper introduces a simple animation system, which is integrated into a web-based solution (JavaScript/HTML5). Since Web browser are commonly available on computers, the presented application is easy to use on any platform from any location (easy to maintain and share). Our system combines the expressive power of web pages for visualising content on-the-fly with a fully fledged interactive (physics-based) animation solution that includes a rich set of libraries.