We introduce a benchmark dataset, ROV6D, for 6D pose estimation of remotely operated vehicles (ROVs). The proposed dataset includes the training synthetic data generation and the testing data collection in both a pool and the open lake at Maoming. The training subset consists of a large number of synthetic images with 6D ground-truth poses. The testing subsets, including the Pool subset and Maoming subset, focus on challenging cases with various levels of occlusion and visibility.
Our dataset ROV6D can be found here (BaiduNetDisk, password: brov).