Case Study: Bosch achieves LiDAR semantic segmentation and embedded deployment with MathWorks (MATLAB)

Designing Lidar Sensor Classifier Using Matlab Framework

Bosch needed to develop a LiDAR sensor classifier for semantic segmentation and to build a framework‑independent, interoperable deep‑learning workflow that could be deployed across SoC/cloud and embedded hardware. To tackle LiDAR data ingestion, preprocessing, model training/validation, and automatic target code generation, Bosch used MathWorks' MATLAB framework (including Model constructors and MATLAB Coder) with support from the MathWorks engineering team.

MathWorks delivered an end‑to‑end MATLAB solution — from data import and subsampling through training/validation, ONNX compatibility, and automated C/C++ code generation — enabling Bosch to run inference on x86 and embedded targets. The MathWorks workflow produced successful code generation and on‑target segmentation (reconstructed organized point clouds, confusion matrices, and validated performance), speeding layer standardization, model fine‑tuning, and multi‑target deployment of embedded binaries and executables.

Bosch