Quantum of Vision: the speed vs accuracy tradeoff in visual recognition
Speed and accuracy are both important for visual systems. Biological or artificial, visual systems often need to trade off the accuracy of their decisions versus the amount of time spent on collecting information. Examples include animals avoiding predators and autonomous vehicles driving at night. What is the optimal strategy for this tradeoff? Is the human visual system optimal and if so, how does the system implement the optimal strategy? Can we implement this strategy in practical computer vision applications? Come find out.