Deep Learning for Lidar

Apply artificial intelligence techniques to lidar applications

Train, test, and deploy deep learning networks on lidar point clouds for object detection and semantic segmentation.

With MATLAB and Simulink, you can:

Preprocess lidar point clouds for applying deep learning algorithms
Use the Lidar Labeler app to label lidar point clouds for object detection
Handle large amounts of data for training, testing, and validation with datastores
Generate C/C++ and CUDA codes for deep learning workflows for semantic segmentation and object detection on point cloud data

Why Use Deep Learning for Lidar?

Aerial lidar point cloud segmented based on objects like building, vegetation, vehicles, and more.

Lidar Semantic Segmentation

Apply deep learning algorithms to segment lidar point clouds. Train, test, and evaluate semantic segmentation networks, including PointNet++, PointSeg, and SqueezeSegV2, on lidar data.

Segmentation

Detect cars and trucks from point cloud data and fit oriented bounding box around them.

Object Detection on Lidar Point Clouds

Detect and fit oriented bounding boxes around objects in lidar point clouds and use them for object tracking or lidar labeling workflows. Design, train, and evaluate robust detectors such as PointPillars networks.

Detection and Tracking

Lidar Labeling

Label lidar point clouds for training deep learning models. Apply built-in or custom algorithms to automate lidar point cloud labeling with the Lidar Labeler app and evaluate automation algorithm performance.

Lidar Labeling

Segmented point cloud showing cars and background.

Deployment

Generate CUDA^® MEX code for networks like PointPillars, SqueezeSegV2, and PointNet++ to deploy point cloud segmentation or object detection algorithms on GPUs.

Code Generation for Lidar Object Detection