In other words: "Objects with less surface area and lower albedo are less reliably picked up by visual deep neural nets. We haven't benchmarked this against hum

In other words: "Objects with less surface area and lower albedo are less reliably picked up by visual deep neural nets. We haven't benchmarked this against human drivers."