[16:11 Sun,28.July 2019 by Thomas Richter] |
The next algorithm ("https://repo-sam.inria.fr/fungraph/deep-relighting/ (Multi-view Relighting using a Geometry-Aware Network)") makes a similar thing possible for whole objects: ;) Stonehenge interactively relightened for a time-lapse simulation
Large geometric objects in videos can thus be illuminated retrospectively in a controlled manner - for example, the lighting of a video can be made more dramatic - for example, the drone video of a Mexican pyramid. The neural network uses geometric clues from several images from different perspectives of the object to produce a realistic shadow cast under different lighting conditions. For example, a time-lapse effect can be simulated from a simple drone video by a rapidly changing shadow cast - analogous to the course of the sun over the period of a normal time-lapse video. The script programmed by the researchers allows the user to define the beginning and end position of the sun&s course with two clicks or to show the sun&s position and the resulting shadow cast by means of an image.  more infos at bei repo-sam.inria.frdeutsche Version dieser Seite: Lichtstimmung in Drohnen-Videos von Bauwerken nachträglich ändern // Siggraph 2019 |
 |
|
|