Data for ZF-AIM

# Zebra finch data release ``` zf_data_release/ ├── Raw stereo recordings: *.flac ├── zebra_finch_metadata.xlsx ├── processed_data/ # Aggregated pipeline outputs and analysis │ ├── call_latencies_<condition>_<id>.csv # Per-call latency and feature tables, one per pair │ ├── OneSided_Map.csv, Natural_SideMap.xlsx # Bird ID / side mappings │ └── ZFAIM_Revision1_ToUpload.R # Statistical analysis script ├── zebra_finch_data_pipeline/ # Processing pipeline code (see its README) ├── voxaboxen_data/ # Voxaboxen detector training data (audio + Raven selection tables) │ ├── train_data/ │ └── test_data/ ├── voxaboxen_weights/moredata1/ # Trained voxaboxen detector (model.pt, params.yaml) and test-set outputs └── bird_mixit_model_checkpoints/ # BirdMixIT separation checkpoints (Google Research, Apache 2.0) ``` ## Code - Pipeline: `zebra_finch_data_pipeline/`, also at https://github.com/earthspecies/zebra_finch_data_pipeline - Voxaboxen (detector training/inference): https://github.com/nkundiushuti/voxaboxen - BirdMixIT: https://github.com/google-research/sound-separation/blob/master/models/bird_mixit/README.md ## Data - Raw audio: Four example raw audio tracks from recordings with two birds (Pair27), the yoked control (Pair3003), ZF-AIM (Pair 5014), and ZF-AIM-ablate (Pair5031). - Processed data: Collection of csv's produced by running `zebra_finch_data_pipeline` on raw audio. - Voxaboxen data: data for training and evaluating the ZF-Voxaboxen sound event detection model ## Model weights - Voxaboxen: Model weights for zebra finch detection and diarization in stereo recordings. - Bird Mixit: Model weights from Google's Mixit model trained for separating bird vocalizations. Mixit model weights are provided under an Apache license, with citation in the corresponding directory. These data are provided under a CC-BY-NC license. Voxaboxen model weights and code are provided under a MIT license. If you use either, please cite: *James, L.S., Hoffman, B., Liu, J.Y., Miron, M., Alizadeh, M., Fernandez, E., Geist, M., Kim, D., Raskin, A., Sakata, J.T., Chemla, E., Pietquin, O., and Woolley, S. 2026. Zebra finch females flexibly communicate with each other and with AI-driven acoustic interaction models.*