A new method developed by FIT researchers will contribute to more realistic videogame and virtual reality environments

A new method developed by researchers from the CPhoto@FIT group can measure the realism of synthetically generated trees. They described the underlying research together with their colleagues from the Purdue University in a paper which was accepted at SIGGRAPH - the most prestigious conference on graphics, which will take place in Tokyo in December. "It is a huge success. We conducted an extensive user study in which 4,000 participants compared over one million pairs of images so that we know which images look the most realistic.  The scores were used to train our neural network, called ICTree. The program is now able to recognise which synthetically generated trees are likely or unlikely to be considered realistic," describes Martin Čadík, one of the authors of the paper. The method can be used by developers of simulators, games, architectural visualisations or virtual reality applications, i.e. wherever synthetically generated 3D trees are used, to design a more realistic environment. More information and video available HERE.