Getting Started with Lidar Labeler App, Part 4: Labeling Automation Using Pretrained Deep Learning Models

Deep learning Automation Algorithms are advanced artificial intelligence algorithms that empower machines to learn and enhance their performance in tasks that traditionally require human intelligence. These applications are particularly effective in computer vision and natural language processing applications, where extensive data processing is necessary. The Lidar labeler app from MATLAB incorporates a powerful deep learning automation algorithm, which enables users to leverage pre-trained deep learning models or import their own custom trained models for object detection in 3D point clouds.

This algorithm supports both Cuboid and Semantic label outputs, providing flexibility and accuracy in labeling. With this Automation Algorithm, users can seamlessly map the network output to the defined label definitions, streamlining the labeling process and achieving precise results. Upon detecting objects, it is recommended to review and manually adjust the labels as necessary, ensuring conformity with the automatically drawn labels. This step ensures the highest level of accuracy and control in the labeling process.