In what appears to be Apple's first publicly disclosed paper on autonomous vehicle technology, we learn that the Cupertino giant has a better way to spot cyclists and pedestrians on the road.
The paper was submitted last week to an independent online journal by Yin Zhou and Oncel Tuzel, who proposed an improved way for helping computers detect three-dimensional objects, reports Autonews.
It's called "VoxelNet" and has apparently produced "highly encouraging results" in spotting pedestrians and cyclists with just Lidar data. The pair also disclosed that they were able to beat other approaches for detecting three-dimensional objects that used only Lidar.
Usually, autonomous cars use a combination of two-dimensional cameras and deep-sensing 'Lidar' units in order to navigate their surroundings. Even though these units provide good depth information, their low resolution makes it difficult to detect small objects in the distance without help from a normal camera. In other words, Apple has apparently found a way to extract more performance from 'Lidar' technology.
It's important to point out that these tests were simulated on a computer and did not involve any cars hitting the road.
Renderings by Aristomenis Tsirbas