Harvesting natural points-of-view preferences for arbitrary objects

Automatically identification of an optimal, representative or relevant viewpoint for a given 3D object is an important task with several applications in digital marketing, visualization, 3D data management and shape retrieval. An objective function to optimize the viewpoint is difficult to define in purely geometric terms, given the semantic and aesthetic bias a user is likely to introduce when presented with a target object. Therefore, supervised data-driven approaches are a natural candidate to address the problem. This implies the availability of large datasets describing natural point-of-view preferences for a number of object categories. In this work, we present a framework designed to harvest this kind a dataset. The system involves tasking end-users with answering general qualitative questions about a displayed 3D object, while silently collecting point-of-view information. In other terms, we capture viewpoint preferences as a latent objective, allowing for an unbiased observation of user behavior in performing various tasks that involve object visualization. We publicly release both our dataset of collected viewpoint statistics and the tool to harvest them, which is easily configurable by experimenters to collect more data in the future or integrate the approach into online 3D viewers.