Measures whether a photograph of an animal can retrieve that animal's call. Each query is a wildlife photograph, and the model ranks 100 candidate taxa -- the photographed taxon plus 99 distractors -- over an index of 1,024 field recordings, where a taxon is represented by every recording of that taxon. A taxon scores its best-matching recording and the 100 taxa are ranked by that score, so the reported… See the full description on the dataset page:
https://huggingface.co/datasets/myang333/BioVITAI2ARetrieval.