Just Bird Model¶
The Just Bird model has a single class, bird. It finds bird vocalizations of any species and draws a tight box around each call. Detections are merged with window fusion by default, which also joins same-band boxes less than 0.05 s apart into one phrase box (--phrase-gap, 0 keeps one box per call).
The current model was released on 20 October 2026 and replaces the March 2026 model.
More on Just-Bird:
- Training: the training recipe and why each step is there, training curves and boxes per split.
- Label Refinement: how the bout labels of four datasets were split into call labels.
- Model Types: the export formats and how closely they match the PyTorch model.
Training Data¶
The model was trained on all five regional datasets and on Bird-Sounds-Global:
- Southwestern Amazon Basin
- Island of Hawai'i
- Northeastern United States
- Southern Sierra Nevada mountain range
- Western United States
- Bird-Sounds-Global: annotated soundscapes from xeno-canto (Lifeplan and SystemShift projects) from many countries
Labels. Hawai'i and Southern Sierra Nevada are annotated per call. The other datasets mostly put one box around a whole song or bout. Their boxes were replaced by single-call boxes in two rounds of self-training: a model trained on call-level labels found the calls inside each human box. A clip with a human box it could not replace was discarded, together with all its other boxes. Training shows why, and Just-Bird Label Refinement how.
Recipe. YOLO11s, started from the COCO weights [14], trained for 10 epochs on PCEN spectrograms. The learning rate decays to its final value within those 10 epochs, and mixup augmentation [31] is switched off for the last 3 (mosaic augmentation is not used). There is no early stopping.
Performance on an Unseen Site¶
Because the model was trained on every dataset, none of them is left to test it on recordings from a site it has not seen. The transfer was therefore measured with the same recipe trained without Southern Sierra Nevada and then run on the full Southern Sierra Nevada recordings. This is a test on a domain the model never saw, like the camera-trap test of the WILDS benchmark, where models are evaluated on camera locations that are not in their training data [11]. The previous model's recipe was retrained and tested in the same way.
Event F1 on full recordings with window fusion (including the removal of nested boxes, phrase gap off):
| Recipe | IoU 0.25 | IoU 0.5 | IoU 0.75 | IoU 0.5 at the threshold chosen on Hawai'i val |
|---|---|---|---|---|
| Current | 0.540 | 0.385 | 0.116 | 0.384 |
| Previous (March 2026) | 0.463 | 0.278 | 0.059 | 0.249 |
The graph shows precision, recall and F1 at IoU 0.5 for every confidence threshold. The curves come from the model trained without Southern Sierra Nevada, not from the released weights, which were trained on it. The dashed line is the previous recipe retrained the same way, not the March 2026 model.
- Confidence threshold: 0.15, the default in
detect_birds.pyand the Demo. The best threshold on the Hawai'i validation recordings is 0.13 to 0.15, and it carries over to the unseen site with almost no loss (0.384 against the best 0.385). - Phrase gap: some Southern Sierra Nevada labels box a whole phrase where the model boxes single syllables, which lowers the scores at IoU 0.5 and above. The default phrase gap of 0.05 s raises F1 at IoU 0.5 there from 0.385 to 0.396, and on song-labelled Western-US validation minutes from 0.170 to 0.197. On call-level labels it costs about 0.05 (Hawai'i val: 0.581 to 0.531). If your annotations box single calls, pass
--phrase-gap 0. Phrase boxes has the numbers for other gaps. - In-domain: on the full Hawai'i validation recordings, the released model reaches F1 0.590 at IoU 0.5 (threshold 0.14, phrase gap off).
References¶
[11] Koh, P. W., Sagawa, S., Marklund, H., et al. (2021). "WILDS: A benchmark of in-the-wild distribution shifts." Proceedings of the 38th International Conference on Machine Learning (ICML), PMLR 139:5637–5664.
[14] Lin, T.-Y., Maire, M., Belongie, S., et al. (2014). "Microsoft COCO: Common objects in context." Computer Vision – ECCV 2014, 740–755.
[31] Zhang, H., Cisse, M., Dauphin, Y. N., & Lopez-Paz, D. (2018). "mixup: Beyond empirical risk minimization." International Conference on Learning Representations (ICLR).