I can see how this would be an improvement for a robotics application, but this is a regression for any type of video enhancement task (fake bokeh, stereoscopic coversion, etc). One would want the depth of the objects being reflected, and not the reflective object. Is this tunable somehow?
I can see how this would be an improvement for a robotics application, but this is a regression for any type of video enhancement task (fake bokeh, stereoscopic coversion, etc). One would want the depth of the objects being reflected, and not the reflective object. Is this tunable somehow?