Data generation toolkit for image based rendering algorithms
V.V. Krishna, P. J. Narayanan · 2006
Computer vision algorithms require accurate and high quality data for experimentation, tuning, and testing. Each research group used to have their own test data initially. Some of the test data came to be shared, making comparison of performance of different algorithms possible. In this paper, we present DGTk a tool for generating data sets primarily for computer vision and image based rendering algorithms. Researchers can compose a scene of their choice using standard models, place lights and cameras in it and compute a variety of representations about the scene. The data supported include images, depth maps, calibration matrices, correspondences etc. The tool provides an intuitive interface for setting up 3D scenes by importing some standard 3D model les. It provides an easy to use interface for setting up and imaging rich static and dynamic scenes and enables the sharing of data among researchers. DGTk goes beyond being a graphics authoring system and is an independent, lightweight, and extensible tool that is computer vision aware.