With just the low dynamic range fixed framerate monocular camera input? I’m not sure I could. Also, everyone knows computers still aren’t as good at vision recognition as humans are.