In this era of advanced technological development, 3D Deep Learning (DL) finds crucial applications in many domains including robotics, autonomous driving, virtual reality, medical diagnosis and so on. 3D point cloud annotation is best suitable for precise object detection.

What is a point cloud?

A point cloud is a group of data points in space that represents a three-dimensional shape with XYZ Cartesian coordinates. Each group of dots represents a section of physical space, thus generating a three-dimensional model. The level of detail increases with point density.

How is point cloud created?

Using 3D sensors like LiDAR (Light Detection and Ranging) or photogrammetry software, the point cloud data is produced. The laser light is emitted from the sensor's source, strikes the target, and then is reflected back. The sensor determines the distance by measuring the time it takes for each pulse to return (Time of Flight). Each of these metrics is converted into a "Point Cloud," a 3D visualization.

3D Cuboid Annotation

3D boxes are used to detect and track various objects in the scene. 3D boxes gives additional depth information about the object.

Point Cloud Segmentation Annotation

Segmentation is a process of classifying an object having additional attributes. Image segmentation simply means partitioning of a digital image for a computer to easily analyze and interpret. In autonomous vehicles, this technique is used to distinguish different types of lanes in 3D point cloud maps with more precise visibility using 3D orientation for safe driving.