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.
VocalSound-C (VS-C) serves as a benchmark for DHAuDS in the context of test-time adaptation for audio classification. VS-C, a subset of VocalSound, utilizes only test-set samples to address domain shift. Each sample in VS-C has a duration of 10 seconds, a sample rate of 16 kHz, and belongs to one of 6 classes. The dataset comprises two sets: adaptation and evaluation sets. Each set comprises seven corruption categories and two levels. Seven corruption categories are defined: WHN, ENQ, END1, END2, ENSC, TST, and PSH. Two levels are defined: L1 and L2, where L2 indicates a higher degree of complexity. As for evaluation and adaptation sets, each comprises 50,274 samples. In total, VS-C consists of 100,548 samples, with each category-level, such as WHN-L1 or WHN-L2, containing 3,591 samples.
- Downloads last month
- 63