Swarm-Based Identification of Animation Key Points from 2D-medialness Maps

Abstract

In this article we present the use of dispersive flies optimisation (DFO) for swarms of particles active on a medialness map-a 2D field representation of shape informed by perception studies. Optimising swarms activity permits to efficiently identify shape-based keypoints to automatically annotate movement and is capable of producing meaningful qualitative descriptions for animation applications. When taken together as a set, these keypoints represent the full body pose of a character in each processed frame. In addition, such keypoints can be used to embody the notion of the Line of Action (LoA), a well known classic technique from the Disney studios used to capture the overall pose of a character to be fleshed out. Keypoints along a medialness ridge are local peaks which are efficiently localised using DFO driven swarms. DFO is optimised in a way so that it does not need to scan every image pixel and always tend to converge at these peaks. A series of experimental trials on different animation characters in movement sequences confirms the promising performance of the optimiser over a simpler, currently-in-use brute-force approach.

Beer Organoleptic Optimisation: Utilising Swarm Intelligence and Evolutionary Computation Methods

“…It was shown that DFO is more efficient in 85% and more reliable in 90% of the 28 standard optimisation benchmarks used; furthermore, when there exists a statistically significant difference, DFO converges to better solutions in 71% of problem set. Furthermore, DFO has been applied to various problems, including but not limited to medical imaging [2], optimising machine learning algorithms [5,6], training deep neural networks for false alarm detection in intensive care units [29], computer vision and quantifying symmetrical complexities [4], identifying animation key points from medialness maps [8] and analysis of autopoiesis in computational creativity [3].…”

Evolutionary Optimisation of Beer Organoleptic Properties: A Simulation Framework

“…It has been demonstrated that DFO has outperformed these algorithms and is used as the optimiser in this work. This algorithm belongs to the broad family of swarm intelligence and evolutionary computation techniques and has been applied to a diverse set of problems including: medical imaging [23], solving diophantine equations [24], PID speed control of DC motor [25], optimising machine learning algorithms [26,27], training deep neural networks for false alarm detection in intensive care units [28], computer vision and quantifying symmetrical complexities [29], identifying animation key points from medialness maps [30], and the analysis of autopoiesis in computational creativity [31].…”

Swarm Optimised Few-View Binary Tomography

“…ϕ is invariably set to 1 and ∆ to 0.001 in published studies (e.g. [3,6,31,7,5]). The upper bound on ϕ is derived from a convergence analysis for stochastic difference equations [11]).…”