config.json

{
  "model_type": "audio-classification",
  "architecture": "resnet18",
  "task": "underwater-acoustic-classification",
  "num_classes": 4,
  "class_names": [
    "vessel",
    "marine_animal",
    "natural_sound",
    "other_anthropogenic"
  ],
  "class_to_id": {
    "vessel": 0,
    "marine_animal": 1,
    "natural_sound": 2,
    "other_anthropogenic": 3
  },
  "audio_config": {
    "sample_rate": 16000,
    "duration": 10.0,
    "n_mels": 128,
    "n_fft": 2048,
    "hop_length": 512,
    "fmax": 8000
  },
  "model_config": {
    "input_channels": 1,
    "pretrained_backbone": "imagenet",
    "frozen_layers": ["conv1", "layer1"],
    "trainable_layers": ["layer2", "layer3", "layer4", "fc"]
  },
  "training_info": {
    "epochs": 4,
    "batch_size": 8,
    "learning_rate": 0.0001,
    "optimizer": "Adam",
    "loss_function": "CrossEntropyLoss",
    "accuracy": 0.9833,
    "balanced_accuracy": 0.9167
  },
  "framework": "pytorch",
  "pytorch_version": "2.0+",
  "license": "mit"
}