dam_features_scores_dynamic_range.metadata.json

{
  "source_model": "KintsugiHealth/dam",
  "model_dir": "dam",
  "input": {
    "name": "whisper_logmel_features",
    "dtype": "float32",
    "shape": [
      1,
      80,
      3000
    ],
    "description": "Preprocessed DAM/Whisper log-mel features, not raw audio."
  },
  "output": {
    "dtype": "float32",
    "shape": [
      2
    ],
    "labels": [
      "depression_score",
      "anxiety_score"
    ]
  },
  "preprocessing": {
    "sample_rate_hz": 16000,
    "chunk_seconds": 30,
    "n_mels": 80,
    "n_frames": 3000,
    "notes": [
      "Load mono audio and resample to 16 kHz.",
      "Remove DC offset and normalize amplitude to [-1, 1].",
      "Pad/split audio into 30-second chunks.",
      "Run Whisper log-mel feature extraction as in DAM featex.py.",
      "Apply DAM logmel_energies rescale before inference."
    ]
  },
  "thresholds": {
    "depression": [
      -0.6699,
      -0.2908
    ],
    "anxiety": [
      -0.7939,
      -0.2173,
      0.1521
    ]
  },
  "raw_score_order": [
    "depression",
    "anxiety"
  ]
}