Published June 6, 2026 | Version v1
Dataset Open

Multi-Species Animal Vocalization Raw Dataset for CNN-Transformer Dual Branch Recognition

Authors/Creators

  • 1. ROR icon Beijing University of Posts and Telecommunications
  • 2. BUPT

Description

Open dataset supporting IEEE paper on dual-branch CNN-Transformer animal species & emotion classification.

Dataset includes 8 animal categories (cat, dog, pig, sheep, goat, horse, cow, turkey), fine-grained emotion labels only for cat and dog.

Data split rule: 20% fixed independent test set, remaining 80% for training & 10-fold stratified cross-validation to optimize α_sp/α_emo fusion weights.

Files

dataset.zip

Files (963.6 MB)

Name Size
md5:ca45d26f58199c0598aecb52cbfea21b
963.6 MB Preview Download