Combining complementary object recognition methods lets computers see better