Case Study: Drass Group achieves real-time maritime object detection with MathWorks

Drass Develops Deep Learning System for Real-Time Object Detection in Maritime Environments

Drass Group, a maritime technology company, needed a way to help ship operators monitor sea environments and detect objects, obstacles, and other ships in real time. Because maritime scenes are difficult to analyze and there were no pretrained object-detection models for this environment, the team had to build its own deep learning solution from scratch. MathWorks provided the MATLAB tools they used to create, train, test, and validate the model.

Using MATLAB Deep Learning Toolbox, Image Processing Toolbox, Video Labeler, Parallel Computing Toolbox, and GPU Coder, Drass Group built a YOLOv2-based object detection system, automated data labeling, and deployed the model to NVIDIA GPUs on ships as C++ code. MathWorks helped the team reduce labeling time from an estimated 249 hours to 42 hours for 5,000 frames and complete the project in 10 months, two months ahead of deadline, while establishing a flexible framework for future model updates and retraining.