V33DA++

V33DA++: Overlapping-Call and Long-Context Companion to V33DA V33DA++ is the companion release to the V33DA benchmark. It is built from the same BirdPark recordings and the same 10 zebra finches, but covers data outside the strict single-vocalizer attribution clips that define V33DA. V33DA++ is intended for tasks that benefit from overlapping callers or longer temporal context: source separation, vocal activity detection, audio-visual synchronization, and active-speaker detection.… See the full description on the dataset page: pp.

Source: Hugging Face Hub (songbirdini/v33da_pp). Metadata imported from the dataset’s Hub tags.