Talks

Perception

Thursday, 14 August 3:45 PM - 5:15 PM | Vancouver Convention Centre, West Building, Rooms 118-120 Session Chair: Jeremy Kenisky, Geomedia, Inc.

High-Level Saliency Prediction for Smart Game Balancing

The first automated high-level saliency predictor that incorporates the schema and singleton hypotheses into the differential-weighting model. The predictor estimates the probabilities of objects to be attended and facilitates game-prop placement. Game difficulty is adjusted, since topology affects object-search completion time.

George Koulieris
Technical University Of Crete

George Drettakis
REVES/INRIA Sophia-Antipolis

Douglas Cunningham
Brandenburgische Technische Universität

Katerina Mania
Technical University of Crete

Perceptually Based Parameter Adjustments for Video-Processing Operations

A method that formulates adjustment of the parameters of image and video processing operators as an interactive optimization problem based on the user's assessment as the objective function to rapidly find the perceptually best settings, overcoming the often difficult and time-consuming tweaking process of complex operators.

Gabriel Eilertsen
Linköpings universitet

Jonas Unger
Linköpings universitet

Temporally Coherent Video De-Anaglyph

This approach to convert existing anaglyph videos into stereoscopic 3D videos with full-color views builds on a temporal-coherence framework that also produces temporally coherent disparity maps and optical flow fields, which can be used for various post-processing tasks.

Joan Sol Roo
Inria

Christian Richardt
Inria

OpenVL: A Developer-Level Abstraction of Computer Vision

OpenVL is a developer-centred abstraction for computer vision, designed to be easy to use for general developers and researchers in other fields to integrate vision techniques in their applications without requiring significant knowledge of or expertise in computer vision.

Gregor Miller
The University of British Columbia

Sidney Fels
The University of British Columbia