{
  "scene_id": "room",
  "scene_label": "Room",
  "scene_type": "indoor",
  "dataset": "Mip-NeRF 360",
  "dataset_page": "https://jonbarron.info/mipnerf360/",
  "pretrained_model_source": "Graphdeco official 3D Gaussian Splatting pretrained-model release, previously downloaded to this workspace.",
  "model_project_page": "https://github.com/graphdeco-inria/gaussian-splatting",
  "model_release_page": "https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/",
  "model_license_page": "https://github.com/graphdeco-inria/gaussian-splatting/blob/main/LICENSE.md",
  "license_status": "Consult the upstream dataset and Graphdeco license terms before redistribution or commercial use.",
  "reuse_method": "NTFS hard links for source images, COLMAP and model payloads; small camera/split metadata files are copied independently to isolate later edits.",
  "local_source": "D:\\3DGSData\\route_a\\01_sources\\mipnerf360\\room",
  "local_model": "D:\\3DGSData\\route_a\\02_pretrained\\room",
  "registered_cameras": 311,
  "images_subdir": "images_2",
  "colmap_directory": "sparse/0",
  "existing_working_manifest": {
    "scene_id": "room",
    "scene_type": "indoor",
    "dataset": "mipnerf360",
    "source_path": "D:\\3DGSData\\route_a\\01_sources\\mipnerf360\\room",
    "images_subdir": "images_2",
    "registered_cameras": 311,
    "train_cameras": 272,
    "test_cameras": 39,
    "camera_format": "Graphdeco-compatible c2w; cx/cy added",
    "model_status": "pretrained",
    "preprocessing_status": "ready for current stage"
  },
  "pretrained_model": {
    "iteration": 30000,
    "ply": "point_cloud/iteration_30000/point_cloud.ply",
    "gaussian_count": 1593376,
    "training_status": "previously downloaded and locally verified; no retraining performed in this task"
  },
  "delivery_metadata": {
    "photo_count": 311,
    "camera_count": 311,
    "scene_bytes": 809777992,
    "preview_files": [
      "previews/preview_01.jpg",
      "previews/preview_02.jpg",
      "previews/preview_03.jpg",
      "previews/preview_04.jpg"
    ],
    "preview_selection": "four evenly spaced images selected only from train.txt; test images were not used",
    "known_limitations": [
      "Pretrained public release; the exact upstream training command may not have been provided.",
      "Inspect boundary floaters, reflective/transparent surfaces and thin geometry before freezing experimental ROIs."
    ],
    "suggested_instance_rois": [
      "??",
      "??",
      "????"
    ],
    "reference_training_command": "python train.py -s . -m <output_dir> --eval --iterations 30000",
    "qa_scope": "automated structural validation; suggested ROIs require project-team visual confirmation"
  }
}