{
  "status": "estimated",
  "method": "instance-count-stability-v1.2",
  "model_source": "mask2former",
  "frames_sampled": 16,
  "frames": [
    0,
    8,
    16,
    24,
    32,
    40,
    48,
    56,
    64,
    72,
    80,
    88,
    96,
    104,
    112,
    120
  ],
  "thresholds": {
    "min_transitions": 5,
    "min_count_range": 3,
    "person_score": 0.9,
    "person_min_frames": 2,
    "singleton_min_frames": 2,
    "popin_presence_frac": 0.4,
    "popin_score": 0.95,
    "popin_min_appearances": 3,
    "min_presence_frames": 2,
    "hands_score": 0.8,
    "hands_min_frames": 2,
    "max_hands": 2
  },
  "singleton_classes": [
    "dining table",
    "microwave",
    "oven",
    "refrigerator",
    "sink"
  ],
  "classes": {
    "banana": {
      "counts": [
        0,
        1,
        1,
        0,
        2,
        2,
        2,
        1,
        2,
        2,
        2,
        3,
        1,
        2,
        2,
        1
      ],
      "violations": []
    },
    "bowl": {
      "counts": [
        0,
        0,
        0,
        0,
        0,
        0,
        1,
        0,
        0,
        0,
        1,
        0,
        0,
        0,
        0,
        0
      ],
      "violations": []
    },
    "counter": {
      "counts": [
        0,
        0,
        0,
        0,
        1,
        1,
        0,
        1,
        0,
        0,
        0,
        0,
        0,
        0,
        0,
        0
      ],
      "violations": []
    },
    "person": {
      "counts": [
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1,
        1
      ],
      "violations": []
    }
  },
  "ignored_classes": [
    "light",
    "wall-other-merged"
  ],
  "violations": [],
  "verdict": "pass",
  "confidence": "low \u2014 v1.2 heuristic screener; VLM adherence audit remains not_configured",
  "coverage_note": "COCO panoptic vocabulary has no 'faucet' and no 'water' class: faucet duplication and fluid rigidity are NOT directly detectable. This screener checks instance-count stability only over detectable classes (person, sink, oven, refrigerator, microwave, dining table, cup, bowl, bottle, banana, apple, knife, ...). When hand_poses.jsonl exists, a MediaPipe hands_anomaly signal additionally covers phantom third hands.",
  "hands": {
    "source": "hand_poses.jsonl",
    "frames_sampled": 16,
    "frames": [
      0,
      8,
      16,
      24,
      32,
      40,
      48,
      56,
      64,
      72,
      80,
      88,
      96,
      104,
      112,
      120
    ],
    "confident_hands_per_frame": [
      1,
      0,
      1,
      1,
      1,
      2,
      1,
      1,
      1,
      0,
      0,
      0,
      0,
      1,
      1,
      2
    ]
  }
}