Wolfset: A High-Quality Underwater Acoustic Dataset for Algorithm Development and Analysis

As data becomes increasingly available, relying on quality datasets for algorithm analysis and development is essential. However, data gathering can be expensive and time-consuming, and this process must be optimized to allow others to reuse data with simplicity and accuracy. The Wolfset is an acoustic dataset gathered using a Bruel & Kjaer type 8104 hydrophone in an anechoic tank usually used for ships' sonar calibration. The name Wolfset is inspired by the Seawolf submarine class, renowned for its advanced sound source detection and classification capabilities. Using an anechoic tank, we can obtain a high-quality dataset representing acoustic sources without undesired external perturbations. In many operating conditions, several outboard motors and an electric motor from a basic remotely controlled ship model were used as sound sources, usually called targets. Then, external transients and noise sources were added to approximate the dataset to the sounds present in real-world conditions. This dataset uses a systematic approach to demonstrate the diversity and accuracy needed for effective algorithm development.