Evaluating Motion Estimation Models from Behavioural and Psychophysical Data

Offering proper evaluation methodology is essential to continue progress in modelling the neural mechanisms involved in vision information processing. Currently the evaluation of biologically inspired motion estimation models lacks a proper methodology for comparing their performance against behavioural and psychophysical data. Here we set the basis for such a new benchmark methodology based on human visual performance and designed a database of image sequences taken from neuroscience and psychophysics literature. In this article we focused on two fundamental aspects of motion estimation, which are the respective influence between d versus d cues and the dynamics of motion integration. Since motion models deal with many kinds of motion representations and scales, we defined two general readouts based on a global motion estimation. Such readouts, namely eye movements and perceived motion, will serve as a reference to compare simulated and experimental data. Baseline results are provided for biologically inspired artificial vision models but also for computer vision models. As a whole we provide here the basis for a valuable evaluation methodology to unravel the fundamental mechanisms of motion perception in the visual cortex. Our database is freely available on the web together with scoring instructions and results at: http://www-sop.inria.fr/neuromathcomp/psymotionbench