DHAuDS: A Dynamic and Heterogeneous Audio Benchmark for Test-Time Adaptation
Paper
• 2511.18421 • Published
Dataset Viewer
The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
ReefSet-C (RS-C) serves as a benchmark for DHAuDS in the context of test-time adaptation for audio classification. RS-C, a subset of ReefSet, utilizes only test-set samples to address domain shift. Each sample in RS-C has a duration of 1.88 seconds, a sample rate of 16 kHz, and belongs to one of 37 classes. The dataset comprises seven corruption sets: WHN, ENQ, END1, END2, ENSC, TST, and PSH. Each corruption set is divided into two levels, L1 and L2, where L2 indicates a higher degree of complexity. RS-C is further divided into evaluation and adaptation sets, each comprising 239,918 samples and including both levels of all seven corruptions. In total, RS-C consists of 479,836 samples, with each corruption level, such as WHN-L1 or WHN-L2, containing 17,137 samples.
- Downloads last month
- 4