3 ms·
It sounds like you may not have experience in this domain? You're benchmarking an active, time-based task (a person "seeing a car" - implicitly allowing eye sa
by mturmon 9y ago
It sounds like you may not have experience in this domain? You're benchmarking an active, time-based task (a person "seeing a car" - implicitly allowing eye saccade, multiple images, context, depth) versus a static task ("classify this single image based on no context"). They are not comparable.
- ouid 9y agohave you actually tried the imagenet task yourself? The reason imagenet is popular is because it has allowed google to claim that they can beat human performance on image recognition by artifically lowering human performance on image recognition.