whale-call-detector
This model is a fine-tuned version of MIT/ast-finetuned-audioset-10-10-0.4593 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.3449
- Accuracy: 0.8866
- Precision: 0.8880
- Recall: 0.8866
- F1: 0.8670
- F1 Water: 0.8421
- F1 Resident: 0.9167
- F1 Transient: 0.8696
- F1 Humpback: 0.8148
- F1 Vessel: 0.9091
- F1 Jingle: 0.8
- F1 Human: 0.9412
- F1 Bird: 0.9565
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 3e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 24
- num_epochs: 5
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1 | F1 Water | F1 Resident | F1 Transient | F1 Humpback | F1 Vessel | F1 Jingle | F1 Human | F1 Bird |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.7617 | 1.0 | 49 | 0.5305 | 0.8247 | 0.8498 | 0.8247 | 0.7535 | 0.9 | 0.7778 | 0.7 | 0.7826 | 0.8571 | 0.75 | 0.9412 | 0.9565 |
| 0.2555 | 2.0 | 98 | 0.5517 | 0.8763 | 0.8928 | 0.8763 | 0.8910 | 0.7143 | 0.875 | 0.9091 | 0.8889 | 0.8333 | 0.8421 | 0.9412 | 0.9565 |
| 0.0240 | 3.0 | 147 | 0.3540 | 0.9072 | 0.9147 | 0.9072 | 0.9181 | 0.8 | 0.9167 | 0.9091 | 0.9286 | 0.9524 | 0.8 | 0.9412 | 0.9565 |
| 0.0016 | 4.0 | 196 | 0.3271 | 0.8969 | 0.8995 | 0.8969 | 0.8846 | 0.8421 | 0.9565 | 0.8696 | 0.8276 | 0.9091 | 0.8 | 0.9412 | 0.9565 |
| 0.0010 | 5.0 | 245 | 0.3462 | 0.8866 | 0.8880 | 0.8866 | 0.8670 | 0.8421 | 0.9167 | 0.8696 | 0.8148 | 0.9091 | 0.8 | 0.9412 | 0.9565 |
Framework versions
- Transformers 5.19.0
- Pytorch 2.14.1+cu130
- Datasets 5.1.0
- Tokenizers 0.23.3