Abstract:
The EISATS website offers various video data for testing stereo and motion analysis algorithms. Due to the increasing use of those data by the international computer vision community, it appeared to be of benefit to offer sources for 3D visualization of stereo an motion analysis results, especially for cases where no ground truth is available for evaluating the accuracy of results. This MSc project contributes to the task to verify how particular stereo or motion analysis algorithms perform on those sequences if only subjective means (i.e. no ground truth) are available. For this purpose, analysis results are shown in 3D space. The visualization is based on OpenGL and on specifically designed colour keys for disparity and optic flow data.
