examples/training_sample/processed/00000002/train_reference_instances.json

{
  "target_video": "train_target_rgb.mp4",
  "prompts": [
    "person",
    "radio",
    "chessboard",
    "food box"
  ],
  "instances": [
    {
      "rank": 0,
      "obj_id": 2,
      "max_area_ratio": 0.21596304086538462,
      "frame_indices": [
        7,
        47,
        48
      ]
    },
    {
      "rank": 1,
      "obj_id": 1,
      "max_area_ratio": 0.17424128605769232,
      "frame_indices": [
        4,
        32,
        53
      ]
    },
    {
      "rank": 2,
      "obj_id": 3,
      "max_area_ratio": 0.13515875400641025,
      "frame_indices": [
        16,
        22,
        55
      ]
    },
    {
      "rank": 3,
      "obj_id": 4,
      "max_area_ratio": 0.13495843349358974,
      "frame_indices": [
        2,
        10,
        53
      ]
    },
    {
      "rank": 4,
      "obj_id": 5,
      "max_area_ratio": 0.029286858974358974,
      "frame_indices": [
        44,
        45,
        64
      ]
    }
  ]
}