Case Study: OCIUS Technology Limited improves 3D underwater labeling accuracy with Segments.ai

How an autonomous underwater robot is turning noisy point cloud data into crystal-clear 3D locations

OCIUS Technology Limited, a maritime robotics company, faced the challenge of evaluating the performance of its autonomous underwater vehicle's object detection algorithms. The system fused data from multiple sensors, but the underwater environment produced low-resolution, noisy point cloud data, making it difficult to establish a reliable ground truth for comparison. To solve this, they turned to the vendor Segments and its multi-sensor data labeling platform.

Segments.ai provided a solution with an interface that allowed Ocius to simultaneously view and label fused 2D image and 3D point cloud data. This overlay of perspectives was essential for creating accurate manual labels in all six degrees of freedom from the noisy data. Using Segments.ai, Ocius was able to generate high-quality datasets, export the results, and effectively compare the algorithm's automated detections against the ground truth, which streamlined debugging and improved the reliability of their autonomous systems.

OCIUS Technology Limited

Sarina H.

Robotics Software Engineer