Semi-Permanent Real-Time Oceanic Sensing Platform for Innovation Challenges

Today's great achievements in Artificial Intelligence are largely due to large annotated datasets made available more than a decade ago. Unfortunately, not all research fields benefit from such data access. Underwater acoustics is one of those fields, despite the need for more researchers tackling public-domain pressing issues. A few underwater acoustic datasets exist on the internet for detecting, tracking and localizing ships, marine mammals, etc. However, none of those cover the same area under a variety of conditions for extended periods with target truth-data and recorded environmental changes, thus limiting possibilities for studying robustness topics. This article presents a multi-organizational initiative that will help correct the situation. Indeed, a real-time oceanic platform is built to collect complete datasets from Halifax Harbor for more than one year. This will constitute a unique, and one of the largest, public-domain dataset providing complete environmental, acoustic, and ship-traffic conditions. Halifax ferries are tagged with GPS transmitters to act as ground truth. Detecting, tracking and localizing noise-emitting sources using underwater acoustics is notoriously difficult in shallowwater and busy harbour environments. To facilitate this task, the paper contributes by: explaining key technical details of the platform and environment, providing a short overview of factors impacting underwater acoustic propagation, and presenting an analysis of preliminary data collected in August 2024. Data processing identifies significant differences at different times of the day. In addition, a few relevant facts have been quantified.