For the past few years, Apple has been rumored to be working on its own self-driving car. The company has finally admitted that it has been working on autonomous tech and is sharing its progress in a recently released paper.
Apple AI and machine learning researchers Yin Zhou and Oncel Tuzel published a paper last week detailing the company’s progress with its self-driving project. Even though the company has yet to test its vehicle on the streets, it does have many people believing that Apple is raising the standard for self-driving vehicles.
The writers of the paper have devised what they call VoxelNet, which is an architecture for detecting small obstacles using the Light Detection and Ranging (LiDAR) sensing method. They note that VoxelNet is better than the LiDAR-based systems at spotting not only cars but pedestrians and cyclists.
VoxelNet divides a point cloud into equally spaced 3D voxels and transforms a group of points within each voxel into a unified feature representation through the newly introduced voxel feature encoding (VFE) layer. In this way, the point cloud is encoded as a descriptive volumetric representation, which is then connected to a RPN to generate detections. Experiments on the KITTI car detection benchmark show that VoxelNet outperforms the state-of-the-art LiDAR based 3D detection methods by a large margin. Furthermore, our network learns an effective discriminative representation of objects with various geometries, leading to encouraging results in 3D detection of pedestrians and cyclists, based on only LiDAR.
In the past, Apple has been spotted testing LiDR-equipped SUVs on the roads of California and had even begun trialing self-driving short-haul shuttles between its campuses earlier in the year.
We’ll see if Apple decides to share more of its progress in the future on this project, but for now, if you’re interested in reading the full paper, you can click here to read the PDF.