Update jumbotron and ego dataset metadata

Adds the synced RPX jumbotron video, ego view manifests/previews, Dataset Viewer configs, and rerun/download helper updates.

README.md

@@ -32,11 +32,21 @@ configs:
     path: "splits/medium.parquet"
   - split: "hard"
     path: "splits/hard.parquet"
-- config_name: media_preview
+- config_name: mos_phase_preview
   description: "Data Studio media table with one video-layout RGB/depth/mask image preview per MOS phase."
   data_files:
   - split: "preview"
-    path: "preview/media_preview.parquet"
+    path: "preview/mos_phase_preview.parquet"
+- config_name: ego_frames
+  description: "Frame-level manifest rows for optional egocentric MOS views. Ego does not contribute to ESD."
+  data_files:
+  - split: "ego"
+    path: "manifest/ego_frames_v1.parquet"
+- config_name: ego_preview
+  description: "Data Studio media table for egocentric MOS RGB/depth/mask/fisheye previews."
+  data_files:
+  - split: "preview"
+    path: "preview/ego_preview.parquet"
 - config_name: single_object
   description: "SOS selected-object catalog; one 360 collection per object, no difficulty split."
   data_files:
@@ -51,25 +61,28 @@ embodied deployment conditions.
 
 * Code: https://github.com/IRVLUTD/RPX
 
-![RPX teaser](assets/rpx_teaser.png)
-
-[Watch the RPX teaser video](assets/rpx-jumbotron.webm)
+<video controls muted loop playsinline width="100%">
+  <source src="assets/rpx-jumbotron.webm" type="video/webm">
+</video>
 
 ## Dataset at a glance
 
 | | |
 |---|---|
-| Multi-object scenes (MOS) | **100** (`scene001` to `scene100`, 3 phases each: clutter / interaction / clean) |
+| Multi-object scenes (MOS) | **100** (`scene001` to `scene100`, 3 ESD phases each: clutter / interaction / clean) |
+| Egocentric MOS views | Optional `scenes/<scene_id>/ego/` captures tracked separately from ESD |
 | Single-object scenes (SOS) | **70 selected objects** (one 360 collection per object) |
-| Frame manifest rows | **110,000** (75,000 MOS + 35,000 selected SOS) |
+| Frame manifest rows | **110,000** in `frames_v1` (75,000 MOS + 35,000 selected SOS); `frames_v2` adds optional ego rows |
 | MOS mask-object rows | **2,100** local mask IDs mapped to global object IDs |
 
 Scene renames are documented in `manifest/scene_name_mapping_v1.csv`,
 mapping original scene names to `scene001` through `scene100`.
 
 The Hugging Face Dataset Viewer exposes `multi_object` as structured split
-tables and includes a `media_preview` config with an actual image media column:
-one video-layout RGB, depth, and mask preview still for each MOS phase.
+tables, `mos_phase_preview` as a Data Studio image table for normal MOS
+phases, `ego_frames` as the separate frame manifest for egocentric captures,
+and `ego_preview` as a separate Data Studio image table for egocentric MOS
+views when those captures are present.
 
 ## Effort-Stratified Difficulty (ESD)
 
@@ -88,6 +101,8 @@ The released split files expose two levels:
 
 Because scene tiers are aggregated, a hard scene can still contain an easy
 phase. Use the phase-level split when downloading or evaluating MOS tasks.
+Egocentric `ego` views are auxiliary MOS captures and do **not** contribute to
+ESD scores, scene tiers, or easy/medium/hard split membership.
 
 The ESD feature weights are fully accounted for in
 `splits/scene_splits.json`: 27 feature names, 27 weights, no missing weights,
@@ -112,9 +127,10 @@ truth for the score/tier assignments and the feature/weight provenance.
 
 ## Modality inventory
 
-The table below describes this cleaned release. `cam_pose_icp` is not included in this cleaned release.
-Use `manifest/frames_v1.parquet` and the identity manifests under `manifest/`
-as the source of truth.
+The table below describes this cleaned release. `cam_pose_icp` is not included
+in this cleaned release. Use `manifest/frames_v1.parquet` and the identity
+manifests under `manifest/` as the source of truth for the v1 MOS/SOS release;
+use `manifest/frames_v2.parquet` when optional ego rows are present.
 
 | modality | files | bytes |
 |---|---:|---:|
@@ -151,19 +167,37 @@ A subsequent call for a different task on the same split (e.g.
 `relative_pose`) reuses the cached RGB tars and only fetches the new
 modality (`cam_pose`) as the delta.
 
+Egocentric views are not difficulty-split. Until `rpx_benchmark` exposes the
+same contract directly, download them through the Hub allow-pattern helper:
+
+```bash
+python rerun_check/download_ego.py \
+    --repo-id IRVLUTD/RPX \
+    --scenes scene002 \
+    --include-preview
+```
+
+The intended benchmark API contract for the package is a view-aware call such
+as `download_for_task(task="ego_segmentation", split="ego", ...)` or
+`download_for_task(task="segmentation", view="ego", split="ego", ...)`.
+
 ## Repo layout
 
 ```
 IRVLUTD/RPX/
 ├── manifest/
-│   ├── frames_v1.parquet                 # per-frame metadata
+│   ├── frames_v1.parquet                 # v1 per-frame metadata for MOS phases + SOS
+│   ├── frames_v2.parquet                 # v2 metadata including optional ego rows
+│   ├── ego_frames_v1.csv/parquet         # frame metadata for scenes/<scene_id>/ego
 │   ├── scene_name_mapping_v1.csv         # original scene names to scene001..scene100
 │   ���── selected_sos_objects_v1.csv       # selected 70-object SOS catalog
 │   ├── selected_sos_objects_v1.parquet   # Dataset Viewer SOS catalog table
 │   ├── object_catalog_v1.json            # SOS object/global-ID catalog
 │   ├── mos_raw_mask_object_map_v1.csv    # MOS local mask IDs from sam2 metadata
 │   ├── mos_mask_object_map_v1.csv        # MOS local mask IDs joined to SOS global IDs
-│   ├── mos_mask_object_map_v1.parquet    # parquet copy of the MOS map
+│   ├── mos_mask_object_map_v1.parquet    # parquet copy of the MOS phase map
+│   ├── mos_ego_mask_object_map_v1.csv    # ego local mask IDs joined to global IDs
+│   ├── mos_ego_mask_object_map_v1.parquet
 │   └── current.json                      # default version per label modality
 ├── splits/
 │   ├── scene_splits.json
@@ -171,10 +205,16 @@ IRVLUTD/RPX/
 │   ├── easy.csv  medium.csv  hard.csv      # human-readable split tables
 │   └── easy.parquet  medium.parquet  hard.parquet  # Dataset Viewer split tables
 ├── preview/
-│   ├── media_preview.parquet               # Dataset Viewer image media table
-│   ├── data_studio_preview.csv             # preview index with source shard links
+│   ├── mos_phase_preview.parquet           # Dataset Viewer image media table for MOS phases
+│   ├── mos_phase_preview.csv               # source index for MOS phase previews
+│   ├── ego_preview.parquet                 # Dataset Viewer image media table for ego views
+│   ├── ego_preview.csv                     # source index for ego previews
+│   ├── data_studio_preview.csv             # legacy name for the MOS phase preview index
 │   └── image_examples/preview/             # source JPEGs for the media preview
-├── scenes/<scene_id>/<phase>/                     # MOS
+├── scenes/<scene_id>/<phase>/                     # MOS ESD phases: 0, 1, 2
+│   ├── rgb.tar  depth.tar  fisheye.tar
+│   └── labels/{cam_pose,masks,masks_aux,sam2_meta}/v1.tar
+├── scenes/<scene_id>/ego/                         # optional egocentric MOS view
 │   ├── rgb.tar  depth.tar  fisheye.tar
 │   └── labels/{cam_pose,masks,masks_aux,sam2_meta}/v1.tar
 ├── objects/<object_id>/0/                         # SOS
@@ -213,6 +253,11 @@ That row means `scene001/0` mask ID `2` is `boot.2`, whose global object ID
 is `11`, with questionnaire
 `objects_meta/boot.2/questionnaire.json` and SOS template `objects/boot.2/`.
 
+Ego masks follow the same local-ID rule. Use
+`manifest/mos_ego_mask_object_map_v1.csv` or parquet to join
+`scenes/<scene_id>/ego/labels/masks/v1.tar` local mask IDs to global object
+IDs. Ego local IDs are scoped to `scene_id + ego`.
+
 ## Tasks
 
 ### Multi-object (use a difficulty split)
@@ -236,6 +281,15 @@ is `11`, with questionnaire
 | `object_templates_rgbd` | ['depth', 'rgb'] → ['masks'] |
 | `object_pose_library` | ['depth', 'rgb'] → ['cam_pose', 'masks'] |
 
+### Egocentric MOS views (no ESD split)
+
+| recipe | inputs → labels |
+|---|---|
+| `ego_segmentation` | ['rgb'] → ['masks'] |
+| `ego_rgbd_segmentation` | ['depth', 'rgb'] → ['masks'] |
+| `ego_relative_pose` | ['rgb'] → ['cam_pose'] |
+| `ego_stereo_depth` | ['fisheye'] → ['depth'] |
+
 ## Label versioning
 
 Labels live at `labels/<name>/v<N>.tar`. Newer versions land at new
@@ -252,7 +306,7 @@ To pin to a specific version:
 
 ```python
 download_for_task(
-    task="relative_pose", split="easy", repo_id="itaykadosh/RPX",
+    task="relative_pose", split="easy", repo_id="IRVLUTD/RPX",
     label_versions={"cam_pose": "v1"},   # don't auto-upgrade to v2
 )
 ```
@@ -265,7 +319,7 @@ download_for_task(
               embodied perception},
     author = {IRVL UT Dallas},
     year   = 2026,
-    url    = {https://huggingface.co/datasets/itaykadosh/RPX},
+    url    = {https://huggingface.co/datasets/IRVLUTD/RPX},
 }
 ```
 

assets/rpx-jumbotron.webm

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5f75d0941c77b0c1737de3b0541a839828174c8465990ea3bd2794e5118869fc
-size 18241914
+oid sha256:78116823231cc577c4ca987b902f4c3c7faee22ca5dc3c322202aad934e2b527
+size 5640054

ego_release/README.md

@@ -0,0 +1,68 @@
+# RPX Ego Release Prep
+
+This folder contains the staging script for the upcoming egocentric MOS view.
+It does not recompute ESD. Ego is treated as an auxiliary view under each MOS
+scene:
+
+```text
+scenes/<scene_id>/ego/
+```
+
+## Expected Source Shape
+
+The script searches for capture directories with this shape:
+
+```text
+<ego_source>/<scene_number>/<capture>/
+├── rgb/*.png
+├── depth/*.png
+├── fisheye/left/*.png
+├── fisheye/right/*.png
+├── cam_pose/*.npz
+└── sam2/
+    ├── masks/*.png
+    ├── mask_to_object.json
+    └── ...
+```
+
+The first numeric ancestor is mapped through
+`manifest/scene_name_mapping_v1.csv`, so `ego_example/2/2` becomes
+`scenes/scene002/ego`.
+
+## Generate Upload-Ready Files
+
+From the repo root:
+
+```bash
+python -m venv /tmp/rpx-ego-venv
+/tmp/rpx-ego-venv/bin/pip install -r rerun_check/requirements.txt huggingface_hub
+/tmp/rpx-ego-venv/bin/python ego_release/prepare_ego_release.py \
+    --ego-source-root ego_example \
+    --apply \
+    --overwrite
+```
+
+The generated HF-facing files are:
+
+- `scenes/<scene_id>/ego/*.tar`
+- `manifest/ego_frames_v1.csv`
+- `manifest/ego_frames_v1.parquet`
+- `manifest/frames_v2.parquet`
+- `manifest/mos_ego_mask_object_map_v1.csv`
+- `manifest/mos_ego_mask_object_map_v1.parquet`
+- `preview/ego_preview.csv`
+- `preview/ego_preview.parquet`
+- `preview/mos_phase_preview.parquet`
+
+## Verify
+
+```bash
+/tmp/rpx-ego-venv/bin/python rerun_check/check_all_modalities.py \
+    --kinds mos,sos,ego
+```
+
+Open the generated Rerun recording:
+
+```bash
+/tmp/rpx-ego-venv/bin/rerun rerun_check/out_all/rpx_all_modalities_visual_check.rrd
+```

ego_release/ego_release_summary.json

@@ -0,0 +1,22 @@
+{
+  "captures": [
+    {
+      "scene_id": "scene002",
+      "target_root": "scenes/scene002/ego",
+      "shard_counts": {
+        "rgb": 251,
+        "depth": 251,
+        "fisheye": 502,
+        "cam_pose": 251,
+        "masks": 251,
+        "masks_aux": 4761,
+        "sam2_meta": 5
+      }
+    }
+  ],
+  "ego_frame_rows": 251,
+  "frames_v2_rows": 110251,
+  "ego_mask_object_rows": 7,
+  "ego_preview_rows": 1,
+  "ego_contributes_to_esd": false
+}

ego_release/prepare_ego_release.py

@@ -0,0 +1,763 @@
+#!/usr/bin/env python3
+"""Prepare egocentric MOS captures for the RPX Hugging Face dataset layout.
+
+The source tree is expected to contain one or more capture directories with
+siblings such as rgb/, depth/, fisheye/, cam_pose/, and sam2/. Each capture is
+packed into scenes/<scene_id>/ego/ and the HF-facing manifest, preview, and
+parquet files are regenerated.
+"""
+
+from __future__ import annotations
+
+import argparse
+import csv
+import io
+import json
+import shutil
+import tarfile
+from collections import defaultdict
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any
+
+import numpy as np
+from PIL import Image, ImageDraw, ImageFont
+
+
+LABEL_VERSION = "v1"
+EGO_VIEW = "ego"
+RGBD_SIZE = (640, 480)
+FISHEYE_SIZE = (848, 800)
+PHASE_NAMES = {0: "clutter", 1: "interaction", 2: "clean"}
+
+
+@dataclass(frozen=True)
+class Capture:
+    source_root: Path
+    scene_id: str
+    target_root: Path
+    frame_indices: tuple[int, ...]
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument(
+        "--repo-root",
+        type=Path,
+        default=Path(__file__).resolve().parents[1],
+        help="RPX dataset repository root.",
+    )
+    parser.add_argument(
+        "--ego-source-root",
+        type=Path,
+        required=True,
+        help="Root containing raw ego captures, e.g. ego_example.",
+    )
+    parser.add_argument(
+        "--apply",
+        action="store_true",
+        help="Write tars/manifests/previews. Without this, only prints the inferred captures.",
+    )
+    parser.add_argument(
+        "--overwrite",
+        action="store_true",
+        help="Replace an existing scenes/<scene>/ego directory.",
+    )
+    parser.add_argument(
+        "--preview-frame",
+        type=int,
+        default=None,
+        help="Frame index for preview images. Default: middle frame per capture.",
+    )
+    return parser.parse_args()
+
+
+def read_csv_rows(path: Path) -> list[dict[str, str]]:
+    with path.open(newline="", encoding="utf-8") as handle:
+        return list(csv.DictReader(handle))
+
+
+def write_csv_rows(path: Path, rows: list[dict[str, Any]], fieldnames: list[str]) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    with path.open("w", newline="", encoding="utf-8") as handle:
+        writer = csv.DictWriter(handle, fieldnames=fieldnames)
+        writer.writeheader()
+        writer.writerows(rows)
+
+
+def load_json(path: Path) -> Any:
+    return json.loads(path.read_text(encoding="utf-8"))
+
+
+def rel(path: Path, repo_root: Path) -> str:
+    return str(path.relative_to(repo_root))
+
+
+def scene_number_map(repo_root: Path) -> dict[str, str]:
+    rows = read_csv_rows(repo_root / "manifest" / "scene_name_mapping_v1.csv")
+    return {row["scene_number"]: row["new_scene_name"] for row in rows}
+
+
+def infer_scene_id(capture_root: Path, source_root: Path, number_map: dict[str, str]) -> str:
+    parts = capture_root.relative_to(source_root).parts
+    for part in parts:
+        if part.startswith("scene") and part[5:].isdigit():
+            return f"scene{int(part[5:]):03d}"
+        if part.isdigit():
+            return number_map.get(part, f"scene{int(part):03d}")
+    raise ValueError(f"could not infer scene id from {capture_root}")
+
+
+def numeric_files(path: Path, suffix: str) -> list[Path]:
+    return sorted(
+        [item for item in path.glob(f"*{suffix}") if item.stem.isdigit()],
+        key=lambda item: int(item.stem),
+    )
+
+
+def numeric_indices(path: Path, suffix: str) -> tuple[int, ...]:
+    return tuple(int(item.stem) for item in numeric_files(path, suffix))
+
+
+def find_capture_roots(source_root: Path, repo_root: Path) -> list[Capture]:
+    number_map = scene_number_map(repo_root)
+    captures: list[Capture] = []
+    for path in sorted(source_root.rglob("*")):
+        if not path.is_dir():
+            continue
+        if not (path / "rgb").is_dir():
+            continue
+        if not ((path / "depth").is_dir() or (path / "fisheye").is_dir() or (path / "cam_pose").is_dir()):
+            continue
+        frame_indices = numeric_indices(path / "rgb", ".png")
+        if not frame_indices:
+            continue
+        scene_id = infer_scene_id(path, source_root, number_map)
+        captures.append(
+            Capture(
+                source_root=path,
+                scene_id=scene_id,
+                target_root=repo_root / "scenes" / scene_id / EGO_VIEW,
+                frame_indices=frame_indices,
+            )
+        )
+    return captures
+
+
+def add_file_to_tar(tar: tarfile.TarFile, src: Path, arcname: str) -> None:
+    info = tar.gettarinfo(str(src), arcname)
+    info.mtime = 0
+    info.uid = 0
+    info.gid = 0
+    info.uname = ""
+    info.gname = ""
+    with src.open("rb") as handle:
+        tar.addfile(info, handle)
+
+
+def write_tar(tar_path: Path, entries: list[tuple[Path, str]]) -> None:
+    tar_path.parent.mkdir(parents=True, exist_ok=True)
+    with tarfile.open(tar_path, "w") as tar:
+        for src, arcname in sorted(entries, key=lambda item: item[1]):
+            add_file_to_tar(tar, src, arcname)
+
+
+def png_entries(directory: Path, prefix: str) -> list[tuple[Path, str]]:
+    if not directory.exists():
+        return []
+    return [(path, f"{prefix}/{path.name}") for path in numeric_files(directory, ".png")]
+
+
+def npz_entries(directory: Path, prefix: str) -> list[tuple[Path, str]]:
+    if not directory.exists():
+        return []
+    return [(path, f"{prefix}/{path.name}") for path in numeric_files(directory, ".npz")]
+
+
+def all_file_entries(directory: Path, prefix: str) -> list[tuple[Path, str]]:
+    if not directory.exists():
+        return []
+    entries: list[tuple[Path, str]] = []
+    for path in sorted(directory.rglob("*")):
+        if path.is_file():
+            entries.append((path, str(Path(prefix) / path.relative_to(directory))))
+    return entries
+
+
+def pack_capture(capture: Capture, repo_root: Path, overwrite: bool) -> dict[str, Any]:
+    if capture.target_root.exists():
+        if not overwrite:
+            raise FileExistsError(f"{capture.target_root} exists; pass --overwrite to replace it")
+        shutil.rmtree(capture.target_root)
+    capture.target_root.mkdir(parents=True, exist_ok=True)
+
+    source = capture.source_root
+    sam2 = source / "sam2"
+    shard_counts: dict[str, int] = {}
+
+    entries = png_entries(source / "rgb", "rgb")
+    write_tar(capture.target_root / "rgb.tar", entries)
+    shard_counts["rgb"] = len(entries)
+
+    entries = png_entries(source / "depth", "depth")
+    if entries:
+        write_tar(capture.target_root / "depth.tar", entries)
+    shard_counts["depth"] = len(entries)
+
+    fisheye_entries = png_entries(source / "fisheye" / "left", "fisheye/left")
+    fisheye_entries += png_entries(source / "fisheye" / "right", "fisheye/right")
+    if fisheye_entries:
+        write_tar(capture.target_root / "fisheye.tar", fisheye_entries)
+    shard_counts["fisheye"] = len(fisheye_entries)
+
+    entries = npz_entries(source / "cam_pose", "cam_pose")
+    if entries:
+        write_tar(capture.target_root / "labels" / "cam_pose" / f"{LABEL_VERSION}.tar", entries)
+    shard_counts["cam_pose"] = len(entries)
+
+    entries = png_entries(sam2 / "masks", "sam2/masks")
+    if entries:
+        write_tar(capture.target_root / "labels" / "masks" / f"{LABEL_VERSION}.tar", entries)
+    shard_counts["masks"] = len(entries)
+
+    aux_entries: list[tuple[Path, str]] = []
+    if sam2.exists():
+        for child in sorted(sam2.iterdir()):
+            if child.name == "masks" or not child.is_dir():
+                continue
+            aux_entries.extend(all_file_entries(child, f"sam2/{child.name}"))
+    if aux_entries:
+        write_tar(capture.target_root / "labels" / "masks_aux" / f"{LABEL_VERSION}.tar", aux_entries)
+    shard_counts["masks_aux"] = len(aux_entries)
+
+    meta_entries = [(path, f"sam2/{path.name}") for path in sorted(sam2.iterdir()) if path.is_file()] if sam2.exists() else []
+    if meta_entries:
+        write_tar(capture.target_root / "labels" / "sam2_meta" / f"{LABEL_VERSION}.tar", meta_entries)
+    shard_counts["sam2_meta"] = len(meta_entries)
+
+    return {"scene_id": capture.scene_id, "target_root": rel(capture.target_root, repo_root), "shard_counts": shard_counts}
+
+
+def read_split_lookup(repo_root: Path) -> dict[tuple[str, int], dict[str, str]]:
+    lookup: dict[tuple[str, int], dict[str, str]] = {}
+    for split in ("easy", "medium", "hard"):
+        for row in read_csv_rows(repo_root / "splits" / f"{split}.csv"):
+            lookup[(row["scene_id"], int(row["phase_index"]))] = row
+    return lookup
+
+
+def write_parquet(path: Path, rows: list[dict[str, Any]], schema: Any | None = None, metadata: dict[bytes, bytes] | None = None) -> None:
+    import pyarrow as pa
+    import pyarrow.parquet as pq
+
+    path.parent.mkdir(parents=True, exist_ok=True)
+    table = pa.Table.from_pylist(rows, schema=schema)
+    if metadata:
+        table = table.replace_schema_metadata(metadata)
+    pq.write_table(table, path)
+
+
+def ego_frame_schema() -> Any:
+    import pyarrow as pa
+
+    return pa.schema(
+        [
+            ("scene_id", pa.string()),
+            ("scene_type", pa.string()),
+            ("view", pa.string()),
+            ("capture_id", pa.string()),
+            ("frame_idx", pa.int64()),
+            ("frame_filename", pa.string()),
+            ("split", pa.string()),
+            ("has_rgb", pa.bool_()),
+            ("has_depth", pa.bool_()),
+            ("has_fisheye", pa.bool_()),
+            ("has_cam_pose", pa.bool_()),
+            ("has_masks", pa.bool_()),
+            ("has_masks_aux", pa.bool_()),
+            ("has_sam2_meta", pa.bool_()),
+            ("shard_rgb", pa.string()),
+            ("shard_depth", pa.string()),
+            ("shard_fisheye", pa.string()),
+            ("shard_cam_pose", pa.string()),
+            ("shard_masks", pa.string()),
+            ("shard_masks_aux", pa.string()),
+            ("shard_sam2_meta", pa.string()),
+        ]
+    )
+
+
+def ego_mask_map_schema() -> Any:
+    import pyarrow as pa
+
+    return pa.schema(
+        [
+            ("scene_id", pa.string()),
+            ("view", pa.string()),
+            ("capture_id", pa.string()),
+            ("local_mask_id", pa.int64()),
+            ("object_id", pa.string()),
+            ("global_object_id", pa.int64()),
+            ("source_catalog_id", pa.string()),
+            ("object_name", pa.string()),
+            ("class_name", pa.string()),
+            ("questionnaire_path", pa.string()),
+            ("sos_path", pa.string()),
+            ("sos_data_path", pa.string()),
+            ("mask_source", pa.string()),
+        ]
+    )
+
+
+def build_ego_frame_rows(repo_root: Path, captures: list[Capture]) -> list[dict[str, Any]]:
+    rows: list[dict[str, Any]] = []
+    for capture in captures:
+        root = rel(capture.target_root, repo_root)
+        has_depth = (capture.target_root / "depth.tar").exists()
+        has_fisheye = (capture.target_root / "fisheye.tar").exists()
+        has_cam_pose = (capture.target_root / "labels" / "cam_pose" / f"{LABEL_VERSION}.tar").exists()
+        has_masks = (capture.target_root / "labels" / "masks" / f"{LABEL_VERSION}.tar").exists()
+        has_masks_aux = (capture.target_root / "labels" / "masks_aux" / f"{LABEL_VERSION}.tar").exists()
+        has_sam2_meta = (capture.target_root / "labels" / "sam2_meta" / f"{LABEL_VERSION}.tar").exists()
+        for frame_idx in capture.frame_indices:
+            filename = f"{frame_idx:05d}.png"
+            rows.append(
+                {
+                    "scene_id": capture.scene_id,
+                    "scene_type": "multi_object_ego",
+                    "view": EGO_VIEW,
+                    "capture_id": EGO_VIEW,
+                    "frame_idx": frame_idx,
+                    "frame_filename": filename,
+                    "split": EGO_VIEW,
+                    "has_rgb": True,
+                    "has_depth": has_depth,
+                    "has_fisheye": has_fisheye,
+                    "has_cam_pose": has_cam_pose,
+                    "has_masks": has_masks,
+                    "has_masks_aux": has_masks_aux,
+                    "has_sam2_meta": has_sam2_meta,
+                    "shard_rgb": f"{root}/rgb.tar",
+                    "shard_depth": f"{root}/depth.tar" if has_depth else "",
+                    "shard_fisheye": f"{root}/fisheye.tar" if has_fisheye else "",
+                    "shard_cam_pose": f"{root}/labels/cam_pose/{LABEL_VERSION}.tar" if has_cam_pose else "",
+                    "shard_masks": f"{root}/labels/masks/{LABEL_VERSION}.tar" if has_masks else "",
+                    "shard_masks_aux": f"{root}/labels/masks_aux/{LABEL_VERSION}.tar" if has_masks_aux else "",
+                    "shard_sam2_meta": f"{root}/labels/sam2_meta/{LABEL_VERSION}.tar" if has_sam2_meta else "",
+                }
+            )
+    return rows
+
+
+def build_frames_v2_rows(repo_root: Path, ego_rows: list[dict[str, Any]]) -> list[dict[str, Any]]:
+    import pyarrow.parquet as pq
+
+    split_lookup = read_split_lookup(repo_root)
+    old_rows = pq.read_table(repo_root / "manifest" / "frames_v1.parquet").to_pylist()
+    rows: list[dict[str, Any]] = []
+    for row in old_rows:
+        scene_type = row["scene_type"]
+        phase = row["phase"]
+        phase_index = int(phase) if scene_type == "multi_object" else None
+        phase_info = split_lookup.get((row["scene_id"], phase_index), {}) if phase_index is not None else {}
+        capture_id = f"phase{phase_index}" if phase_index is not None else str(phase)
+        rows.append(
+            {
+                **row,
+                "view": "phase" if scene_type == "multi_object" else "object",
+                "capture_id": capture_id,
+                "phase_index": phase_index,
+                "phase_name": PHASE_NAMES.get(phase_index) if phase_index is not None else None,
+                "difficulty": phase_info.get("difficulty"),
+                "rpx_ds": float(phase_info["rpx_ds"]) if phase_info.get("rpx_ds") else None,
+            }
+        )
+    for row in ego_rows:
+        rows.append(
+            {
+                **row,
+                "phase": None,
+                "phase_index": None,
+                "phase_name": None,
+                "difficulty": None,
+                "rpx_ds": None,
+            }
+        )
+    return rows
+
+
+def object_catalog_lookup(repo_root: Path) -> tuple[dict[str, dict[str, Any]], dict[str, list[dict[str, Any]]]]:
+    catalog = load_json(repo_root / "manifest" / "object_catalog_v1.json")
+    by_source = {str(item["source_catalog_id"]): item for item in catalog.get("objects", [])}
+    by_name: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for item in catalog.get("objects", []):
+        by_name[str(item["object_name"])].append(item)
+    return by_source, by_name
+
+
+def read_mask_to_object_from_tar(path: Path) -> dict[str, Any] | None:
+    if not path.exists():
+        return None
+    with tarfile.open(path) as tar:
+        extracted = tar.extractfile("sam2/mask_to_object.json")
+        if extracted is None:
+            return None
+        return json.load(extracted)
+
+
+def build_ego_mask_map_rows(repo_root: Path, captures: list[Capture]) -> list[dict[str, Any]]:
+    by_source, by_name = object_catalog_lookup(repo_root)
+    rows: list[dict[str, Any]] = []
+    for capture in captures:
+        meta_tar = capture.target_root / "labels" / "sam2_meta" / f"{LABEL_VERSION}.tar"
+        mask_to_object = read_mask_to_object_from_tar(meta_tar)
+        if not mask_to_object:
+            continue
+        for mask_id, item in sorted(mask_to_object.items(), key=lambda kv: int(kv[0])):
+            source_catalog_id = str(item["object"]["id"])
+            object_name = str(item["object"]["name"])
+            catalog_item = by_source.get(source_catalog_id)
+            if catalog_item is None:
+                matches = by_name.get(object_name, [])
+                catalog_item = matches[0] if len(matches) == 1 else None
+            if catalog_item is None:
+                raise KeyError(
+                    f"could not join ego mask {capture.scene_id}:{mask_id} "
+                    f"({source_catalog_id}, {object_name}) to object_catalog_v1.json"
+                )
+            rows.append(
+                {
+                    "scene_id": capture.scene_id,
+                    "view": EGO_VIEW,
+                    "capture_id": EGO_VIEW,
+                    "local_mask_id": int(mask_id),
+                    "object_id": catalog_item["object_id"],
+                    "global_object_id": int(catalog_item["global_object_id"]),
+                    "source_catalog_id": str(catalog_item["source_catalog_id"]),
+                    "object_name": catalog_item["object_name"],
+                    "class_name": catalog_item.get("class_name", catalog_item["object_name"]),
+                    "questionnaire_path": catalog_item["questionnaire_path"],
+                    "sos_path": catalog_item.get("sos_path", catalog_item["sos_data_path"]),
+                    "sos_data_path": catalog_item["sos_data_path"],
+                    "mask_source": f"{rel(meta_tar, repo_root)}:sam2/mask_to_object.json",
+                }
+            )
+    return rows
+
+
+def depth_to_uint8(depth: Image.Image) -> Image.Image:
+    arr = np.asarray(depth).astype(np.float32)
+    valid = arr > 0
+    if not valid.any():
+        return Image.fromarray(np.zeros(arr.shape, dtype=np.uint8), mode="L")
+    lo, hi = np.percentile(arr[valid], [2, 98])
+    if hi <= lo:
+        hi = lo + 1
+    norm = np.clip((arr - lo) / (hi - lo), 0, 1)
+    norm[~valid] = 0
+    return Image.fromarray((norm * 255).astype(np.uint8), mode="L")
+
+
+def colorize_mask(mask: Image.Image) -> Image.Image:
+    arr = np.asarray(mask)
+    rgb = np.zeros((*arr.shape, 3), dtype=np.uint8)
+    palette = np.asarray(
+        [
+            [244, 103, 83],
+            [79, 210, 154],
+            [255, 184, 77],
+            [180, 101, 255],
+            [64, 182, 255],
+            [255, 99, 179],
+            [141, 224, 83],
+            [255, 226, 102],
+        ],
+        dtype=np.uint8,
+    )
+    for value in np.unique(arr):
+        if value == 0:
+            continue
+        rgb[arr == value] = palette[(int(value) - 1) % len(palette)]
+    return Image.fromarray(rgb, mode="RGB")
+
+
+def fit_image(image: Image.Image, size: tuple[int, int], nearest: bool = False) -> Image.Image:
+    image = image.convert("RGB")
+    scale = min(size[0] / image.width, size[1] / image.height)
+    new_size = (max(1, round(image.width * scale)), max(1, round(image.height * scale)))
+    method = Image.Resampling.NEAREST if nearest else Image.Resampling.LANCZOS
+    resized = image.resize(new_size, method)
+    canvas = Image.new("RGB", size, (14, 17, 22))
+    canvas.paste(resized, ((size[0] - new_size[0]) // 2, (size[1] - new_size[1]) // 2))
+    return canvas
+
+
+def read_png_from_tar(path: Path, name: str) -> Image.Image | None:
+    if not path.exists():
+        return None
+    with tarfile.open(path) as tar:
+        extracted = tar.extractfile(name)
+        if extracted is None:
+            return None
+        return Image.open(extracted).copy()
+
+
+def load_font(size: int) -> ImageFont.ImageFont:
+    for name in ("Times New Roman.ttf", "Times_New_Roman.ttf", "LiberationSerif-Regular.ttf"):
+        try:
+            return ImageFont.truetype(name, size)
+        except OSError:
+            pass
+    return ImageFont.load_default()
+
+
+def make_ego_preview(capture: Capture, repo_root: Path, frame_idx: int, out_path: Path) -> None:
+    frame = f"{frame_idx:05d}.png"
+    rgb = read_png_from_tar(capture.target_root / "rgb.tar", f"rgb/{frame}")
+    depth = read_png_from_tar(capture.target_root / "depth.tar", f"depth/{frame}")
+    mask = read_png_from_tar(capture.target_root / "labels" / "masks" / f"{LABEL_VERSION}.tar", f"sam2/masks/{frame}")
+    left = read_png_from_tar(capture.target_root / "fisheye.tar", f"fisheye/left/{frame}")
+    right = read_png_from_tar(capture.target_root / "fisheye.tar", f"fisheye/right/{frame}")
+
+    panels = [
+        ("RGB", rgb),
+        ("Depth", depth_to_uint8(depth) if depth is not None else None),
+        ("Masks", colorize_mask(mask) if mask is not None else None),
+        ("Fisheye L", left),
+        ("Fisheye R", right),
+    ]
+    panel_size = (320, 240)
+    label_h = 32
+    footer_h = 46
+    image = Image.new("RGB", (panel_size[0] * len(panels), panel_size[1] + label_h + footer_h), (245, 245, 242))
+    draw = ImageDraw.Draw(image)
+    font = load_font(18)
+    small = load_font(15)
+    for index, (label, panel) in enumerate(panels):
+        x = index * panel_size[0]
+        if panel is None:
+            tile = Image.new("RGB", panel_size, (40, 43, 50))
+            ImageDraw.Draw(tile).text((12, 12), "missing", fill=(235, 235, 235), font=small)
+        else:
+            tile = fit_image(panel, panel_size, nearest=label in {"Depth", "Masks"})
+        image.paste(tile, (x, label_h))
+        draw.text((x + 10, 7), label, fill=(21, 25, 32), font=font)
+    footer_y = label_h + panel_size[1]
+    draw.rectangle((0, footer_y, image.width, image.height), fill=(18, 23, 30))
+    draw.text((14, footer_y + 10), f"RPX EGO / {capture.scene_id} / frame {frame_idx:05d}", fill=(235, 238, 242), font=font)
+    out_path.parent.mkdir(parents=True, exist_ok=True)
+    image.save(out_path, quality=90)
+
+
+def build_ego_preview_rows(repo_root: Path, captures: list[Capture], preview_frame: int | None) -> list[dict[str, Any]]:
+    rows: list[dict[str, Any]] = []
+    for capture in captures:
+        frame_idx = preview_frame if preview_frame is not None else capture.frame_indices[len(capture.frame_indices) // 2]
+        out_name = f"rpx_ego_{capture.scene_id}.jpg"
+        preview_path = repo_root / "preview" / "image_examples" / "ego_preview" / "images" / out_name
+        make_ego_preview(capture, repo_root, frame_idx, preview_path)
+        root = rel(capture.target_root, repo_root)
+        with preview_path.open("rb") as handle:
+            image_bytes = handle.read()
+        rows.append(
+            {
+                "image": {"bytes": image_bytes, "path": out_name},
+                "scene_id": capture.scene_id,
+                "view": EGO_VIEW,
+                "capture_id": EGO_VIEW,
+                "frame_count": len(capture.frame_indices),
+                "preview_frame_index": frame_idx,
+                "source_rgb_shard": f"{root}/rgb.tar",
+                "source_depth_shard": f"{root}/depth.tar",
+                "source_fisheye_shard": f"{root}/fisheye.tar",
+                "source_cam_pose_shard": f"{root}/labels/cam_pose/{LABEL_VERSION}.tar",
+                "source_mask_shard": f"{root}/labels/masks/{LABEL_VERSION}.tar",
+                "source_mask_aux_shard": f"{root}/labels/masks_aux/{LABEL_VERSION}.tar",
+                "source_sam2_meta_shard": f"{root}/labels/sam2_meta/{LABEL_VERSION}.tar",
+                "caption": f"Egocentric RGB-D/mask preview for {capture.scene_id} ({len(capture.frame_indices)} frames).",
+            }
+        )
+    return rows
+
+
+def write_ego_preview_parquet(path: Path, rows: list[dict[str, Any]]) -> None:
+    import pyarrow as pa
+
+    schema = pa.schema(
+        [
+            ("image", pa.struct([("bytes", pa.binary()), ("path", pa.string())])),
+            ("scene_id", pa.string()),
+            ("view", pa.string()),
+            ("capture_id", pa.string()),
+            ("frame_count", pa.int64()),
+            ("preview_frame_index", pa.int64()),
+            ("source_rgb_shard", pa.string()),
+            ("source_depth_shard", pa.string()),
+            ("source_fisheye_shard", pa.string()),
+            ("source_cam_pose_shard", pa.string()),
+            ("source_mask_shard", pa.string()),
+            ("source_mask_aux_shard", pa.string()),
+            ("source_sam2_meta_shard", pa.string()),
+            ("caption", pa.string()),
+        ]
+    )
+    metadata = {
+        b"huggingface": json.dumps(
+            {
+                "info": {
+                    "features": {
+                        "image": {"_type": "Image"},
+                        "scene_id": {"dtype": "string", "_type": "Value"},
+                        "view": {"dtype": "string", "_type": "Value"},
+                        "capture_id": {"dtype": "string", "_type": "Value"},
+                        "frame_count": {"dtype": "int64", "_type": "Value"},
+                        "preview_frame_index": {"dtype": "int64", "_type": "Value"},
+                        "caption": {"dtype": "string", "_type": "Value"},
+                    }
+                }
+            }
+        ).encode("utf-8")
+    }
+    write_parquet(path, rows, schema=schema, metadata=metadata)
+
+
+def write_csv_preview(path: Path, rows: list[dict[str, Any]], repo_id: str = "IRVLUTD/RPX") -> None:
+    csv_rows: list[dict[str, Any]] = []
+    for row in rows:
+        out = {key: value for key, value in row.items() if key != "image"}
+        out["image_preview_url"] = (
+            f"https://huggingface.co/datasets/{repo_id}/resolve/main/"
+            f"preview/image_examples/ego_preview/images/{row['image']['path']}"
+        )
+        csv_rows.append(out)
+    fieldnames = [
+        "scene_id",
+        "view",
+        "capture_id",
+        "frame_count",
+        "preview_frame_index",
+        "image_preview_url",
+        "source_rgb_shard",
+        "source_depth_shard",
+        "source_fisheye_shard",
+        "source_cam_pose_shard",
+        "source_mask_shard",
+        "source_mask_aux_shard",
+        "source_sam2_meta_shard",
+        "caption",
+    ]
+    write_csv_rows(path, csv_rows, fieldnames)
+
+
+def write_current_json(repo_root: Path, ego_rows: list[dict[str, Any]], ego_mask_rows: list[dict[str, Any]]) -> None:
+    path = repo_root / "manifest" / "current.json"
+    current = load_json(path)
+    current["schema_version"] = "v2"
+    current.setdefault("manifests", {}).update(
+        {
+            "frames_v2": "manifest/frames_v2.parquet",
+            "ego_frames": "manifest/ego_frames_v1.parquet",
+            "ego_frames_csv": "manifest/ego_frames_v1.csv",
+            "mos_ego_mask_object_map": "manifest/mos_ego_mask_object_map_v1.parquet",
+            "mos_ego_mask_object_map_csv": "manifest/mos_ego_mask_object_map_v1.csv",
+        }
+    )
+    mos = current.setdefault("mos", {})
+    mos["schema_version"] = "v2"
+    mos["ego_directory_value"] = EGO_VIEW
+    mos["ego_contributes_to_esd"] = False
+    mos["ego_view_count"] = len({row["scene_id"] for row in ego_rows})
+    mos["ego_frame_count"] = len(ego_rows)
+    mos["ego_mask_object_row_count"] = len(ego_mask_rows)
+    mos["ego_frames"] = "manifest/ego_frames_v1.parquet"
+    mos["ego_mask_object_map_csv"] = "manifest/mos_ego_mask_object_map_v1.csv"
+    mos["ego_mask_object_map_parquet"] = "manifest/mos_ego_mask_object_map_v1.parquet"
+    mos["ego_raw_source"] = "scenes/<scene_id>/ego/labels/sam2_meta/v1.tar:sam2/mask_to_object.json"
+    mos["phase_directory_values"] = ["0", "1", "2"]
+    mos["view_directory_values"] = ["0", "1", "2", EGO_VIEW]
+    current.setdefault("metadata_versions", {})["ego_frames"] = "v1"
+    current["metadata_versions"]["mos_ego_mask_object_map"] = "v1"
+    path.write_text(json.dumps(current, indent=2) + "\n", encoding="utf-8")
+
+
+def copy_mos_phase_preview(repo_root: Path) -> None:
+    src = repo_root / "preview" / "media_preview.parquet"
+    dst = repo_root / "preview" / "mos_phase_preview.parquet"
+    if src.exists():
+        shutil.copy2(src, dst)
+    csv_src = repo_root / "preview" / "data_studio_preview.csv"
+    csv_dst = repo_root / "preview" / "mos_phase_preview.csv"
+    if csv_src.exists():
+        shutil.copy2(csv_src, csv_dst)
+
+
+def main() -> None:
+    args = parse_args()
+    repo_root = args.repo_root.resolve()
+    source_root = args.ego_source_root.resolve()
+    captures = find_capture_roots(source_root, repo_root)
+    if not captures:
+        raise SystemExit(f"no ego captures found under {source_root}")
+
+    print("inferred ego captures:")
+    for capture in captures:
+        print(f"  {capture.source_root} -> {capture.target_root} ({len(capture.frame_indices)} frames)")
+
+    if not args.apply:
+        print("dry run only; pass --apply to write dataset artifacts")
+        return
+
+    package_summary = [pack_capture(capture, repo_root, args.overwrite) for capture in captures]
+    ego_frame_rows = build_ego_frame_rows(repo_root, captures)
+    frames_v2_rows = build_frames_v2_rows(repo_root, ego_frame_rows)
+    ego_mask_rows = build_ego_mask_map_rows(repo_root, captures)
+    ego_preview_rows = build_ego_preview_rows(repo_root, captures, args.preview_frame)
+
+    frame_fields = list(ego_frame_rows[0].keys())
+    write_csv_rows(repo_root / "manifest" / "ego_frames_v1.csv", ego_frame_rows, frame_fields)
+    write_parquet(repo_root / "manifest" / "ego_frames_v1.parquet", ego_frame_rows, schema=ego_frame_schema())
+    write_parquet(repo_root / "manifest" / "frames_v2.parquet", frames_v2_rows)
+
+    mask_fields = [
+        "scene_id",
+        "view",
+        "capture_id",
+        "local_mask_id",
+        "object_id",
+        "global_object_id",
+        "source_catalog_id",
+        "object_name",
+        "class_name",
+        "questionnaire_path",
+        "sos_path",
+        "sos_data_path",
+        "mask_source",
+    ]
+    write_csv_rows(repo_root / "manifest" / "mos_ego_mask_object_map_v1.csv", ego_mask_rows, mask_fields)
+    write_parquet(
+        repo_root / "manifest" / "mos_ego_mask_object_map_v1.parquet",
+        ego_mask_rows,
+        schema=ego_mask_map_schema(),
+    )
+
+    write_ego_preview_parquet(repo_root / "preview" / "ego_preview.parquet", ego_preview_rows)
+    write_csv_preview(repo_root / "preview" / "ego_preview.csv", ego_preview_rows)
+    copy_mos_phase_preview(repo_root)
+    write_current_json(repo_root, ego_frame_rows, ego_mask_rows)
+
+    summary = {
+        "captures": package_summary,
+        "ego_frame_rows": len(ego_frame_rows),
+        "frames_v2_rows": len(frames_v2_rows),
+        "ego_mask_object_rows": len(ego_mask_rows),
+        "ego_preview_rows": len(ego_preview_rows),
+        "ego_contributes_to_esd": False,
+    }
+    out_path = repo_root / "ego_release" / "ego_release_summary.json"
+    out_path.write_text(json.dumps(summary, indent=2) + "\n", encoding="utf-8")
+    print(json.dumps(summary, indent=2))
+    print(f"wrote {out_path}")
+
+
+if __name__ == "__main__":
+    main()

manifest/current.json

@@ -5,7 +5,7 @@
     "sam2_meta": "v1",
     "cam_pose": "v1"
   },
-  "schema_version": "v1",
+  "schema_version": "v2",
   "manifests": {
     "frames": "manifest/frames_v1.parquet",
     "scene_name_mapping": "manifest/scene_name_mapping_v1.csv",
@@ -14,7 +14,12 @@
     "mos_raw_mask_object_map": "manifest/mos_raw_mask_object_map_v1.csv",
     "mos_mask_object_map_csv": "manifest/mos_mask_object_map_v1.csv",
     "mos_mask_object_map_parquet": "manifest/mos_mask_object_map_v1.parquet",
-    "mos_mask_object_map": "manifest/mos_mask_object_map_v1.parquet"
+    "mos_mask_object_map": "manifest/mos_mask_object_map_v1.parquet",
+    "frames_v2": "manifest/frames_v2.parquet",
+    "ego_frames": "manifest/ego_frames_v1.parquet",
+    "ego_frames_csv": "manifest/ego_frames_v1.csv",
+    "mos_ego_mask_object_map": "manifest/mos_ego_mask_object_map_v1.parquet",
+    "mos_ego_mask_object_map_csv": "manifest/mos_ego_mask_object_map_v1.csv"
   },
   "sos": {
     "schema_version": "v1",
@@ -26,7 +31,7 @@
     "objects_meta_index": "objects_meta/_index.json"
   },
   "mos": {
-    "schema_version": "v1",
+    "schema_version": "v2",
     "scene_count": 100,
     "phase_count": 300,
     "mask_object_row_count": 2100,
@@ -49,12 +54,29 @@
       "selected_objects": "manifest/selected_sos_objects_v1.csv",
       "object_catalog": "manifest/object_catalog_v1.json",
       "objects_meta_index": "objects_meta/_index.json"
-    }
+    },
+    "ego_directory_value": "ego",
+    "ego_contributes_to_esd": false,
+    "ego_view_count": 1,
+    "ego_frame_count": 251,
+    "ego_mask_object_row_count": 7,
+    "ego_frames": "manifest/ego_frames_v1.parquet",
+    "ego_mask_object_map_csv": "manifest/mos_ego_mask_object_map_v1.csv",
+    "ego_mask_object_map_parquet": "manifest/mos_ego_mask_object_map_v1.parquet",
+    "ego_raw_source": "scenes/<scene_id>/ego/labels/sam2_meta/v1.tar:sam2/mask_to_object.json",
+    "view_directory_values": [
+      "0",
+      "1",
+      "2",
+      "ego"
+    ]
   },
   "metadata_versions": {
     "scene_name_mapping": "v1",
     "selected_sos_objects": "v1",
     "object_catalog": "v1",
-    "mos_mask_object_map": "v1"
+    "mos_mask_object_map": "v1",
+    "ego_frames": "v1",
+    "mos_ego_mask_object_map": "v1"
   }
 }

manifest/ego_frames_v1.csv

@@ -0,0 +1,252 @@
+scene_id,scene_type,view,capture_id,frame_idx,frame_filename,split,has_rgb,has_depth,has_fisheye,has_cam_pose,has_masks,has_masks_aux,has_sam2_meta,shard_rgb,shard_depth,shard_fisheye,shard_cam_pose,shard_masks,shard_masks_aux,shard_sam2_meta
+scene002,multi_object_ego,ego,ego,0,00000.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,1,00001.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,2,00002.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,3,00003.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,4,00004.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,5,00005.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,6,00006.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,7,00007.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,8,00008.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,9,00009.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,10,00010.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,11,00011.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,12,00012.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,13,00013.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,14,00014.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,15,00015.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,16,00016.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,17,00017.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,18,00018.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,19,00019.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,20,00020.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,21,00021.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,22,00022.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,23,00023.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,24,00024.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,25,00025.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,26,00026.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,27,00027.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,28,00028.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,29,00029.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,30,00030.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,31,00031.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,32,00032.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,33,00033.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,34,00034.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,35,00035.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,36,00036.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,37,00037.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,38,00038.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,39,00039.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,40,00040.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,41,00041.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,42,00042.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,43,00043.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,44,00044.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,45,00045.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,46,00046.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,47,00047.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,48,00048.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,49,00049.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,50,00050.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,51,00051.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,52,00052.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,53,00053.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,54,00054.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,55,00055.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,56,00056.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,57,00057.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,58,00058.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,59,00059.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,60,00060.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,61,00061.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,62,00062.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,63,00063.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,64,00064.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,65,00065.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,66,00066.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,67,00067.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,68,00068.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,69,00069.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,70,00070.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,71,00071.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,72,00072.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,73,00073.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,74,00074.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,75,00075.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,76,00076.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,77,00077.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,78,00078.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,79,00079.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,80,00080.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,81,00081.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,82,00082.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,83,00083.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,84,00084.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,85,00085.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,86,00086.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,87,00087.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,88,00088.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,89,00089.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,90,00090.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,91,00091.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,92,00092.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,93,00093.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,94,00094.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,95,00095.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,96,00096.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,97,00097.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,98,00098.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,99,00099.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,100,00100.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,101,00101.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,102,00102.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,103,00103.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,104,00104.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,105,00105.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,106,00106.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,107,00107.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,108,00108.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,109,00109.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,110,00110.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,111,00111.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,112,00112.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,113,00113.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,114,00114.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,115,00115.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,116,00116.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,117,00117.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,118,00118.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,119,00119.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,120,00120.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,121,00121.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,122,00122.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,123,00123.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,124,00124.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,125,00125.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,126,00126.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,127,00127.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,128,00128.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,129,00129.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,130,00130.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,131,00131.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,132,00132.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,133,00133.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,134,00134.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,135,00135.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,136,00136.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,137,00137.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,138,00138.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,139,00139.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,140,00140.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,141,00141.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,142,00142.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,143,00143.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,144,00144.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,145,00145.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,146,00146.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,147,00147.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,148,00148.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,149,00149.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,150,00150.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,151,00151.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,152,00152.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,153,00153.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,154,00154.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,155,00155.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,156,00156.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,157,00157.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,158,00158.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,159,00159.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,160,00160.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,161,00161.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,162,00162.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,163,00163.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,164,00164.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,165,00165.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,166,00166.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,167,00167.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,168,00168.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,169,00169.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,170,00170.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,171,00171.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,172,00172.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,173,00173.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,174,00174.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,175,00175.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,176,00176.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,177,00177.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,178,00178.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,179,00179.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,180,00180.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,181,00181.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,182,00182.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,183,00183.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,184,00184.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,185,00185.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,186,00186.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,187,00187.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,188,00188.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,189,00189.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,190,00190.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,191,00191.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,192,00192.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,193,00193.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,194,00194.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,195,00195.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,196,00196.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,197,00197.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,198,00198.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,199,00199.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,200,00200.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,201,00201.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,202,00202.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,203,00203.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,204,00204.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,205,00205.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,206,00206.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,207,00207.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,208,00208.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,209,00209.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,210,00210.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,211,00211.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,212,00212.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,213,00213.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,214,00214.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,215,00215.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,216,00216.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,217,00217.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,218,00218.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,219,00219.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,220,00220.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,221,00221.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,222,00222.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,223,00223.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,224,00224.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,225,00225.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,226,00226.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,227,00227.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,228,00228.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,229,00229.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,230,00230.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,231,00231.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,232,00232.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,233,00233.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,234,00234.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,235,00235.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,236,00236.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,237,00237.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,238,00238.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,239,00239.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,240,00240.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,241,00241.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,242,00242.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,243,00243.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,244,00244.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,245,00245.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,246,00246.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,247,00247.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,248,00248.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,249,00249.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar
+scene002,multi_object_ego,ego,ego,250,00250.png,ego,True,True,True,True,True,True,True,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar

manifest/ego_frames_v1.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e24c3d1cf171cb266d3ca6fd0b33f33e535cff477c40f49145767a8168430e0
+size 9292

manifest/frames_v2.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8549d200c2d64bf9943f8e97380f265243b33e3dc2521e99d64273f1f718c62c
+size 88949

manifest/mos_ego_mask_object_map_v1.csv

@@ -0,0 +1,8 @@
+scene_id,view,capture_id,local_mask_id,object_id,global_object_id,source_catalog_id,object_name,class_name,questionnaire_path,sos_path,sos_data_path,mask_source
+scene002,ego,ego,1,mug.2,45,88.2,mug,mug,objects_meta/mug.2/questionnaire.json,objects/mug.2/,objects/mug.2/,scenes/scene002/ego/labels/sam2_meta/v1.tar:sam2/mask_to_object.json
+scene002,ego,ego,2,mustard_bottle.3,46,89.3,mustard_bottle,mustard_bottle,objects_meta/mustard_bottle.3/questionnaire.json,objects/mustard_bottle.3/,objects/mustard_bottle.3/,scenes/scene002/ego/labels/sam2_meta/v1.tar:sam2/mask_to_object.json
+scene002,ego,ego,3,cleaning_brush,15,34,cleaning_brush,cleaning_brush,objects_meta/cleaning_brush/questionnaire.json,objects/cleaning_brush/,objects/cleaning_brush/,scenes/scene002/ego/labels/sam2_meta/v1.tar:sam2/mask_to_object.json
+scene002,ego,ego,4,game_controller,23,55,game_controller,game_controller,objects_meta/game_controller/questionnaire.json,objects/game_controller/,objects/game_controller/,scenes/scene002/ego/labels/sam2_meta/v1.tar:sam2/mask_to_object.json
+scene002,ego,ego,5,hammer.2,27,63.2,hammer,hammer,objects_meta/hammer.2/questionnaire.json,objects/hammer.2/,objects/hammer.2/,scenes/scene002/ego/labels/sam2_meta/v1.tar:sam2/mask_to_object.json
+scene002,ego,ego,6,water_bottle.2,70,148.2,water_bottle,water_bottle,objects_meta/water_bottle.2/questionnaire.json,objects/water_bottle.2/,objects/water_bottle.2/,scenes/scene002/ego/labels/sam2_meta/v1.tar:sam2/mask_to_object.json
+scene002,ego,ego,7,remote_controller,52,107,remote_controller,remote_controller,objects_meta/remote_controller/questionnaire.json,objects/remote_controller/,objects/remote_controller/,scenes/scene002/ego/labels/sam2_meta/v1.tar:sam2/mask_to_object.json

manifest/mos_ego_mask_object_map_v1.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1167832ac5d074c16576dd863cc1c0220b8c852f738f16071f5a68b6c5aef393
+size 5163

preview/ego_preview.csv

@@ -0,0 +1,2 @@
+scene_id,view,capture_id,frame_count,preview_frame_index,image_preview_url,source_rgb_shard,source_depth_shard,source_fisheye_shard,source_cam_pose_shard,source_mask_shard,source_mask_aux_shard,source_sam2_meta_shard,caption
+scene002,ego,ego,251,125,https://huggingface.co/datasets/IRVLUTD/RPX/resolve/main/preview/image_examples/ego_preview/images/rpx_ego_scene002.jpg,scenes/scene002/ego/rgb.tar,scenes/scene002/ego/depth.tar,scenes/scene002/ego/fisheye.tar,scenes/scene002/ego/labels/cam_pose/v1.tar,scenes/scene002/ego/labels/masks/v1.tar,scenes/scene002/ego/labels/masks_aux/v1.tar,scenes/scene002/ego/labels/sam2_meta/v1.tar,Egocentric RGB-D/mask preview for scene002 (251 frames).

preview/ego_preview.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8357c93f62448df6b6c20c59971146207f2161ae6fee95aab0a5fc53c4a4b232
+size 76226

preview/mos_phase_preview.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6fca594aa8fb5dd8836d19d3eca3b93d4983d012690ea3cb1d1b4a2524ca29a
+size 44204514

rerun_check/README.md

@@ -0,0 +1,96 @@
+# RPX Rerun Visual Checks
+
+This folder contains visual and structural QA helpers for the RPX release.
+
+## Quick RGB-D Segmentation Check
+
+`make_rerun_check.py` samples one representative middle frame from every scene phase in `preview/data_studio_preview.csv`.
+By default that is all 300 MOS entries: 100 scenes x 3 phases.
+
+It verifies:
+
+- RGB, depth, and mask shards exist.
+- RGB/depth/mask frame counts match.
+- RGB/depth/mask image sizes match.
+- Depth has nonzero values.
+- Masks contain foreground instance IDs.
+
+It writes:
+
+- `out/rpx_visual_check.rrd`: Rerun recording for interactive visual review.
+- `out/rpx_visual_contact_sheet.jpg`: contact sheet for a quick scan.
+- `out/visual_check_report.json`: machine-readable QA report.
+
+This is useful for a fast Data Studio / segmentation preview sanity check, but it is not the full release check.
+
+## Full All-Modality Release Check
+
+`check_all_modalities.py` checks the dataset contract across MOS and SOS:
+
+- 300 MOS scene-phases and 70 SOS selected objects.
+- RGB, depth, masks, fisheye left/right, camera pose, masks_aux, and SAM2 metadata shards.
+- Frame-count alignment and numbered filename sequences.
+- Decoded image samples from each visual stream.
+- Camera pose NPZ fields, finite values, and quaternion norms.
+- MOS mask-object joins against `manifest/mos_mask_object_map_v1.csv`.
+- SOS object metadata and questionnaires.
+- ESD split consistency, Data Studio preview images, release assets, and parquet row counts.
+
+By default it decodes first/middle/last frames for every aligned stream. Use `--sample-policy all` only when you want a much slower exhaustive decode pass.
+
+It writes:
+
+- `out_all/rpx_all_modalities_visual_check.rrd`: Rerun recording for scanning each scene/object unit.
+- `out_all/all_modalities_contact_sheet.jpg`: RGB/depth/mask/fisheye overview for all checked units.
+- `out_all/all_modalities_report.json`: machine-readable QA report.
+- `out_all/all_modalities_summary.md`: short human-readable summary.
+
+## Install
+
+From the repo root:
+
+```bash
+python -m venv /tmp/rpx-rerun-venv
+/tmp/rpx-rerun-venv/bin/pip install -r rerun_check/requirements.txt
+```
+
+If starting from a fresh machine, first use the published RPX Quick Start API:
+
+```bash
+pip install "rpx-benchmark[hub]"
+hf auth login
+python rerun_check/download_rgbd_segmentation.py --repo-id IRVLUTD/RPX
+```
+
+## Run
+
+From the repo root:
+
+```bash
+/tmp/rpx-rerun-venv/bin/python rerun_check/make_rerun_check.py
+```
+
+Run the full all-modality release check:
+
+```bash
+/tmp/rpx-rerun-venv/bin/python rerun_check/check_all_modalities.py
+```
+
+Open the Rerun recording:
+
+```bash
+/tmp/rpx-rerun-venv/bin/rerun rerun_check/out/rpx_visual_check.rrd
+```
+
+Open the full all-modality Rerun recording:
+
+```bash
+/tmp/rpx-rerun-venv/bin/rerun rerun_check/out_all/rpx_all_modalities_visual_check.rrd
+```
+
+For a smaller smoke test:
+
+```bash
+/tmp/rpx-rerun-venv/bin/python rerun_check/make_rerun_check.py --limit 10
+/tmp/rpx-rerun-venv/bin/python rerun_check/check_all_modalities.py --limit 10
+```

rerun_check/check_all_modalities.py

@@ -0,0 +1,1451 @@
+#!/usr/bin/env python3
+"""Full RPX release QA across modalities, labels, manifests, and previews.
+
+This is the broader companion to make_rerun_check.py. It validates all MOS
+scene-phases and all selected SOS objects, not only RGB-D segmentation.
+"""
+
+from __future__ import annotations
+
+import argparse
+import csv
+import io
+import json
+import math
+import tarfile
+from collections import Counter, defaultdict
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any, Iterable
+
+import numpy as np
+from PIL import Image, ImageDraw, ImageFont
+from tqdm import tqdm
+
+
+MOS_EXPECTED_FRAMES = 250
+SOS_EXPECTED_FRAMES = 500
+RGBD_SIZE = (640, 480)
+FISHEYE_SIZE = (848, 800)
+LABEL_VERSION = "v1"
+
+
+@dataclass(frozen=True)
+class Unit:
+    kind: str
+    unit_id: str
+    root: Path
+    expected_frames: int
+    view: str | None = None
+    scene_id: str | None = None
+    phase_index: int | None = None
+    phase_name: str | None = None
+    difficulty: str | None = None
+    object_id: str | None = None
+    global_object_id: int | None = None
+    object_name: str | None = None
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument(
+        "--repo-root",
+        type=Path,
+        default=Path(__file__).resolve().parents[1],
+        help="RPX dataset repository root or downloaded local_dir.",
+    )
+    parser.add_argument("--out-dir", type=Path, default=Path("rerun_check/out_all"))
+    parser.add_argument(
+        "--sample-policy",
+        choices=["middle", "endpoints", "all"],
+        default="endpoints",
+        help=(
+            "Frames to decode inside each stream. 'endpoints' means first/middle/last. "
+            "'all' is exhaustive but can take a long time."
+        ),
+    )
+    parser.add_argument("--limit", type=int, default=None)
+    parser.add_argument(
+        "--kinds",
+        default="mos,sos",
+        help="Comma-separated unit kinds to check: mos, sos, or both.",
+    )
+    parser.add_argument("--image-width", type=int, default=360)
+    parser.add_argument("--contact-thumb-width", type=int, default=140)
+    parser.add_argument("--contact-cols", type=int, default=3)
+    parser.add_argument("--skip-contact-sheet", action="store_true")
+    parser.add_argument("--skip-rerun", action="store_true")
+    parser.add_argument("--fail-on-warnings", action="store_true")
+    return parser.parse_args()
+
+
+def repo_path(repo_root: Path, value: str | Path) -> Path:
+    path = Path(value)
+    return path if path.is_absolute() else repo_root / path
+
+
+def read_csv_rows(path: Path) -> list[dict[str, str]]:
+    with path.open(newline="", encoding="utf-8") as handle:
+        return list(csv.DictReader(handle))
+
+
+def load_json(path: Path) -> Any:
+    return json.loads(path.read_text(encoding="utf-8"))
+
+
+def as_rel(path: Path, repo_root: Path) -> str:
+    try:
+        return str(path.relative_to(repo_root))
+    except ValueError:
+        return str(path)
+
+
+def tar_file_names(path: Path, errors: list[str]) -> list[str]:
+    if not path.exists():
+        errors.append(f"missing shard: {path}")
+        return []
+    try:
+        with tarfile.open(path) as tar:
+            return sorted(member.name for member in tar.getmembers() if member.isfile())
+    except (tarfile.TarError, OSError) as exc:
+        errors.append(f"unreadable tar {path}: {exc}")
+        return []
+
+
+def names_with_prefix(names: Iterable[str], prefix: str, suffix: str) -> list[str]:
+    needle = prefix.rstrip("/") + "/"
+    suffix_lower = suffix.lower()
+    return sorted(
+        name
+        for name in names
+        if name.startswith(needle) and name.lower().endswith(suffix_lower)
+    )
+
+
+def frame_indices_for_count(count: int, policy: str) -> list[int]:
+    if count <= 0:
+        return []
+    if policy == "all":
+        return list(range(count))
+    if policy == "middle":
+        return [count // 2]
+    return sorted({0, count // 2, count - 1})
+
+
+def pick_names(names: list[str], policy: str) -> list[str]:
+    return [names[index] for index in frame_indices_for_count(len(names), policy)]
+
+
+def count_tar_pngs(path: Path, prefix: str) -> int:
+    if not path.exists():
+        return 0
+    try:
+        with tarfile.open(path) as tar:
+            return len(
+                [
+                    member
+                    for member in tar.getmembers()
+                    if member.isfile()
+                    and member.name.startswith(prefix.rstrip("/") + "/")
+                    and member.name.lower().endswith(".png")
+                ]
+            )
+    except (tarfile.TarError, OSError):
+        return 0
+
+
+def numbered_sequence_summary(
+    names: list[str],
+    prefix: str,
+    suffix: str,
+    expected_count: int,
+    errors: list[str],
+    label: str,
+) -> dict[str, Any]:
+    indices: list[int] = []
+    malformed: list[str] = []
+    prefix = prefix.rstrip("/") + "/"
+    for name in names:
+        rel = name[len(prefix) :] if name.startswith(prefix) else name
+        if "/" in rel:
+            malformed.append(name)
+            continue
+        if not rel.lower().endswith(suffix.lower()):
+            malformed.append(name)
+            continue
+        stem = rel[: -len(suffix)]
+        if not stem.isdigit():
+            malformed.append(name)
+            continue
+        indices.append(int(stem))
+
+    count = len(indices)
+    if count != expected_count:
+        errors.append(f"{label} count {count}, expected {expected_count}")
+    expected = set(range(expected_count))
+    actual = set(indices)
+    missing = sorted(expected - actual)
+    extra = sorted(actual - expected)
+    duplicate_count = count - len(actual)
+    if malformed:
+        errors.append(f"{label} has malformed names, first examples: {malformed[:5]}")
+    if missing:
+        errors.append(f"{label} missing frame ids, first examples: {missing[:10]}")
+    if extra:
+        errors.append(f"{label} has unexpected frame ids, first examples: {extra[:10]}")
+    if duplicate_count:
+        errors.append(f"{label} has {duplicate_count} duplicate frame ids")
+
+    return {
+        "count": count,
+        "first": min(indices) if indices else None,
+        "last": max(indices) if indices else None,
+        "missing_count": len(missing),
+        "extra_count": len(extra),
+        "duplicate_count": duplicate_count,
+        "malformed_count": len(malformed),
+    }
+
+
+def image_to_stats(image: Image.Image) -> dict[str, Any]:
+    arr = np.asarray(image)
+    stats: dict[str, Any] = {
+        "mode": image.mode,
+        "size": list(image.size),
+        "dtype": str(arr.dtype),
+    }
+    if arr.size:
+        stats.update(
+            {
+                "min": float(np.min(arr)),
+                "max": float(np.max(arr)),
+                "mean": float(np.mean(arr)),
+                "std": float(np.std(arr)),
+                "nonzero_fraction": float(np.count_nonzero(arr) / arr.size),
+            }
+        )
+    return stats
+
+
+def decode_images(
+    path: Path,
+    sample_names: list[str],
+    errors: list[str],
+    warnings: list[str],
+    label: str,
+    expected_size: tuple[int, int] | None = None,
+    require_nonzero: bool = False,
+) -> tuple[list[dict[str, Any]], dict[str, Image.Image]]:
+    stats: list[dict[str, Any]] = []
+    images: dict[str, Image.Image] = {}
+    if not sample_names:
+        return stats, images
+    try:
+        with tarfile.open(path) as tar:
+            for name in sample_names:
+                try:
+                    extracted = tar.extractfile(name)
+                    if extracted is None:
+                        errors.append(f"{label}: could not extract {name}")
+                        continue
+                    image = Image.open(extracted).copy()
+                except Exception as exc:  # PIL raises several exception types.
+                    errors.append(f"{label}: could not decode {name}: {exc}")
+                    continue
+
+                item = {"name": name, **image_to_stats(image)}
+                if expected_size is not None and image.size != expected_size:
+                    errors.append(
+                        f"{label}: {name} size {image.size}, expected {expected_size}"
+                    )
+                if require_nonzero and item.get("nonzero_fraction", 0.0) <= 0:
+                    warnings.append(f"{label}: {name} has no nonzero pixels")
+                if item.get("std", 0.0) <= 0:
+                    warnings.append(f"{label}: {name} has zero image variance")
+                stats.append(item)
+                images[name] = image
+    except (tarfile.TarError, OSError) as exc:
+        errors.append(f"{label}: could not reopen {path}: {exc}")
+    return stats, images
+
+
+def decode_npz_samples(
+    path: Path,
+    sample_names: list[str],
+    errors: list[str],
+    warnings: list[str],
+    label: str,
+) -> list[dict[str, Any]]:
+    stats: list[dict[str, Any]] = []
+    if not sample_names:
+        return stats
+    try:
+        with tarfile.open(path) as tar:
+            for name in sample_names:
+                try:
+                    extracted = tar.extractfile(name)
+                    if extracted is None:
+                        errors.append(f"{label}: could not extract {name}")
+                        continue
+                    arrays = np.load(io.BytesIO(extracted.read()))
+                except Exception as exc:
+                    errors.append(f"{label}: could not load {name}: {exc}")
+                    continue
+
+                item: dict[str, Any] = {"name": name, "fields": sorted(arrays.files)}
+                if sorted(arrays.files) != ["orientation", "position"]:
+                    errors.append(f"{label}: {name} fields {arrays.files}")
+                for field, shape in [("position", (3,)), ("orientation", (4,))]:
+                    if field not in arrays:
+                        continue
+                    arr = np.asarray(arrays[field])
+                    item[field] = {
+                        "shape": list(arr.shape),
+                        "dtype": str(arr.dtype),
+                        "finite": bool(np.isfinite(arr).all()),
+                        "norm": float(np.linalg.norm(arr)),
+                    }
+                    if arr.shape != shape:
+                        errors.append(f"{label}: {name} {field} shape {arr.shape}")
+                    if not np.isfinite(arr).all():
+                        errors.append(f"{label}: {name} {field} contains non-finite values")
+                if "orientation" in item:
+                    norm = item["orientation"]["norm"]
+                    if not math.isclose(norm, 1.0, rel_tol=0.02, abs_tol=0.02):
+                        warnings.append(f"{label}: {name} quaternion norm is {norm:.4f}")
+                stats.append(item)
+    except (tarfile.TarError, OSError) as exc:
+        errors.append(f"{label}: could not reopen {path}: {exc}")
+    return stats
+
+
+def read_tar_text(tar: tarfile.TarFile, name: str) -> str:
+    extracted = tar.extractfile(name)
+    if extracted is None:
+        raise ValueError(f"could not extract {name}")
+    return extracted.read().decode("utf-8", errors="replace")
+
+
+def read_tar_json(tar: tarfile.TarFile, name: str) -> Any:
+    return json.loads(read_tar_text(tar, name))
+
+
+def depth_to_uint8(depth: Image.Image) -> Image.Image:
+    arr = np.asarray(depth).astype(np.float32)
+    valid = arr > 0
+    if not valid.any():
+        return Image.fromarray(np.zeros(arr.shape, dtype=np.uint8), mode="L")
+    lo, hi = np.percentile(arr[valid], [2, 98])
+    if hi <= lo:
+        hi = lo + 1
+    norm = np.clip((arr - lo) / (hi - lo), 0, 1)
+    norm[~valid] = 0
+    return Image.fromarray((norm * 255).astype(np.uint8), mode="L")
+
+
+def colorize_mask(mask: Image.Image) -> Image.Image:
+    arr = np.asarray(mask)
+    if arr.ndim == 3:
+        return mask.convert("RGB")
+    rgb = np.zeros((*arr.shape, 3), dtype=np.uint8)
+    palette = np.asarray(
+        [
+            [244, 103, 83],
+            [79, 210, 154],
+            [255, 184, 77],
+            [180, 101, 255],
+            [64, 182, 255],
+            [255, 99, 179],
+            [141, 224, 83],
+            [255, 226, 102],
+            [102, 232, 232],
+            [231, 91, 116],
+            [161, 135, 255],
+            [76, 217, 100],
+        ],
+        dtype=np.uint8,
+    )
+    for value in np.unique(arr):
+        if value == 0:
+            continue
+        rgb[arr == value] = palette[(int(value) - 1) % len(palette)]
+    return Image.fromarray(rgb, mode="RGB")
+
+
+def fit_image(image: Image.Image, box: tuple[int, int], *, nearest: bool = False) -> Image.Image:
+    image = image.convert("RGB")
+    max_w, max_h = box
+    scale = min(max_w / image.width, max_h / image.height)
+    size = (max(1, round(image.width * scale)), max(1, round(image.height * scale)))
+    method = Image.Resampling.NEAREST if nearest else Image.Resampling.LANCZOS
+    resized = image.resize(size, method)
+    canvas = Image.new("RGB", box, (16, 18, 22))
+    canvas.paste(resized, ((max_w - size[0]) // 2, (max_h - size[1]) // 2))
+    return canvas
+
+
+def load_font(size: int) -> ImageFont.ImageFont:
+    for name in ("Times New Roman.ttf", "Times_New_Roman.ttf", "LiberationSerif-Regular.ttf"):
+        try:
+            return ImageFont.truetype(name, size)
+        except OSError:
+            pass
+    return ImageFont.load_default()
+
+
+def make_unit_tile(
+    unit: Unit,
+    result: dict[str, Any],
+    viz: dict[str, Image.Image],
+    thumb_width: int,
+) -> Image.Image:
+    panel_h = max(90, round(thumb_width * 0.72))
+    label_h = 38
+    panel_names = ["rgb", "depth", "mask", "fisheye_left", "fisheye_right"]
+    tile = Image.new("RGB", (thumb_width * len(panel_names), panel_h + label_h), (13, 17, 23))
+    draw = ImageDraw.Draw(tile)
+    font = load_font(13)
+    small_font = load_font(11)
+
+    for i, name in enumerate(panel_names):
+        image = viz.get(name)
+        x = i * thumb_width
+        if image is None:
+            panel = Image.new("RGB", (thumb_width, panel_h), (38, 42, 49))
+            ImageDraw.Draw(panel).text((8, 8), "missing", fill=(230, 230, 230), font=small_font)
+        else:
+            nearest = name in {"depth", "mask"}
+            panel = fit_image(image, (thumb_width, panel_h), nearest=nearest)
+        tile.paste(panel, (x, 0))
+        draw.text((x + 5, panel_h - 16), name.replace("_", " "), fill=(235, 235, 235), font=small_font)
+
+    status = result["status"].upper()
+    label = unit.unit_id
+    if unit.kind == "mos":
+        label += f" {unit.difficulty or ''}"
+    elif unit.kind == "ego":
+        label += " ego"
+    else:
+        label += f" gid={unit.global_object_id}"
+    status_color = (83, 210, 154) if result["status"] == "ok" else (255, 184, 77)
+    if result["status"] == "error":
+        status_color = (244, 103, 83)
+    draw.rectangle((0, panel_h, tile.width, tile.height), fill=(245, 245, 242))
+    draw.text((7, panel_h + 5), label[:62], fill=(20, 24, 30), font=font)
+    draw.text((7, panel_h + 21), status, fill=status_color, font=small_font)
+    if result["errors"] or result["warnings"]:
+        msg = (result["errors"] or result["warnings"])[0]
+        draw.text((70, panel_h + 21), msg[:82], fill=(60, 64, 72), font=small_font)
+    return tile
+
+
+def make_contact_sheet(tiles: list[Image.Image], out_path: Path, cols: int) -> None:
+    if not tiles:
+        return
+    cols = max(1, cols)
+    tile_w, tile_h = tiles[0].size
+    rows = (len(tiles) + cols - 1) // cols
+    sheet = Image.new("RGB", (cols * tile_w, rows * tile_h), (245, 245, 242))
+    for index, tile in enumerate(tiles):
+        x = (index % cols) * tile_w
+        y = (index // cols) * tile_h
+        sheet.paste(tile, (x, y))
+    out_path.parent.mkdir(parents=True, exist_ok=True)
+    sheet.save(out_path, quality=88)
+
+
+def open_rerun(rrd_path: Path):
+    import rerun as rr
+
+    rr.init("rpx_all_modalities_visual_check", spawn=False)
+    rr.save(str(rrd_path))
+    return rr
+
+
+def set_rerun_time(rr, index: int) -> None:
+    if hasattr(rr, "set_time_sequence"):
+        rr.set_time_sequence("unit", index)
+    else:
+        rr.set_time("unit", sequence=index)
+
+
+def resize_width(image: Image.Image, width: int, *, nearest: bool = False) -> Image.Image:
+    if image.width <= width:
+        return image.copy()
+    height = max(1, round(image.height * width / image.width))
+    method = Image.Resampling.NEAREST if nearest else Image.Resampling.LANCZOS
+    return image.resize((width, height), method)
+
+
+def log_to_rerun(
+    rr,
+    index: int,
+    unit: Unit,
+    result: dict[str, Any],
+    viz: dict[str, Image.Image],
+    image_width: int,
+) -> None:
+    set_rerun_time(rr, index)
+    if "rgb" in viz:
+        rr.log("rpx/rgb", rr.Image(np.asarray(resize_width(viz["rgb"].convert("RGB"), image_width))))
+    if "depth" in viz:
+        depth = resize_width(viz["depth"], image_width, nearest=True)
+        try:
+            rr.log("rpx/depth", rr.DepthImage(np.asarray(depth), meter=1000.0))
+        except TypeError:
+            rr.log("rpx/depth", rr.DepthImage(np.asarray(depth)))
+    if "mask" in viz:
+        mask = resize_width(viz["mask"], image_width, nearest=True)
+        rr.log("rpx/masks", rr.SegmentationImage(np.asarray(mask)))
+    if "fisheye_left" in viz:
+        left = resize_width(viz["fisheye_left"].convert("RGB"), image_width)
+        rr.log("rpx/fisheye/left", rr.Image(np.asarray(left)))
+    if "fisheye_right" in viz:
+        right = resize_width(viz["fisheye_right"].convert("RGB"), image_width)
+        rr.log("rpx/fisheye/right", rr.Image(np.asarray(right)))
+
+    markdown = [
+        f"## {index:03d} {unit.unit_id}",
+        "",
+        f"- kind: `{unit.kind}`",
+        f"- view: `{unit.view}`",
+        f"- status: `{result['status']}`",
+        f"- expected frames: `{unit.expected_frames}`",
+    ]
+    if unit.kind == "mos":
+        markdown.extend(
+            [
+                f"- scene: `{unit.scene_id}`",
+                f"- phase: `{unit.phase_index}` / `{unit.phase_name}`",
+                f"- difficulty: `{unit.difficulty}`",
+            ]
+        )
+    elif unit.kind == "ego":
+        markdown.extend(
+            [
+                f"- scene: `{unit.scene_id}`",
+                "- ESD: `not applicable`",
+            ]
+        )
+    else:
+        markdown.extend(
+            [
+                f"- object: `{unit.object_id}`",
+                f"- global object id: `{unit.global_object_id}`",
+                f"- object name: `{unit.object_name}`",
+            ]
+        )
+    for message in result["errors"][:5]:
+        markdown.append(f"- error: {message}")
+    for message in result["warnings"][:5]:
+        markdown.append(f"- warning: {message}")
+    if hasattr(rr, "TextDocument"):
+        rr.log("rpx/metadata", rr.TextDocument("\n".join(markdown), media_type="text/markdown"))
+
+
+def build_units(repo_root: Path, kinds: set[str]) -> tuple[list[Unit], dict[str, Any]]:
+    context: dict[str, Any] = {}
+    units: list[Unit] = []
+
+    split_rows: list[dict[str, str]] = []
+    for split in ("easy", "medium", "hard"):
+        csv_path = repo_root / "splits" / f"{split}.csv"
+        rows = read_csv_rows(csv_path)
+        for row in rows:
+            row = dict(row)
+            row["split"] = split
+            split_rows.append(row)
+    context["split_rows"] = split_rows
+
+    if "mos" in kinds:
+        for row in sorted(split_rows, key=lambda r: (r["scene_id"], int(r["phase_index"]))):
+            phase_index = int(row["phase_index"])
+            units.append(
+                Unit(
+                    kind="mos",
+                    unit_id=f"{row['scene_id']}.phase{phase_index}",
+                    root=repo_root / "scenes" / row["scene_id"] / str(phase_index),
+                    expected_frames=MOS_EXPECTED_FRAMES,
+                    view="phase",
+                    scene_id=row["scene_id"],
+                    phase_index=phase_index,
+                    phase_name=row["phase_name"],
+                    difficulty=row["difficulty"],
+                )
+            )
+
+    selected_rows = read_csv_rows(repo_root / "manifest" / "selected_sos_objects_v1.csv")
+    context["selected_sos_rows"] = selected_rows
+    if "sos" in kinds:
+        for row in sorted(selected_rows, key=lambda r: int(r["global_object_id"])):
+            units.append(
+                Unit(
+                    kind="sos",
+                    unit_id=f"object:{row['object_id']}",
+                    root=repo_root / "objects" / row["object_id"] / "0",
+                    expected_frames=SOS_EXPECTED_FRAMES,
+                    view="object",
+                    object_id=row["object_id"],
+                    global_object_id=int(row["global_object_id"]),
+                    object_name=row["object_name"],
+                )
+            )
+
+    ego_rows: list[dict[str, str]] = []
+    ego_frames_csv = repo_root / "manifest" / "ego_frames_v1.csv"
+    if ego_frames_csv.exists():
+        ego_rows = read_csv_rows(ego_frames_csv)
+    context["ego_frame_rows"] = ego_rows
+    if "ego" in kinds:
+        if ego_rows:
+            grouped: dict[str, list[dict[str, str]]] = defaultdict(list)
+            for row in ego_rows:
+                grouped[row["scene_id"]].append(row)
+            for scene_id, rows in sorted(grouped.items()):
+                units.append(
+                    Unit(
+                        kind="ego",
+                        unit_id=f"{scene_id}.ego",
+                        root=repo_root / "scenes" / scene_id / "ego",
+                        expected_frames=len(rows),
+                        view="ego",
+                        scene_id=scene_id,
+                        phase_name="ego",
+                    )
+                )
+        else:
+            for root in sorted((repo_root / "scenes").glob("scene*/ego")):
+                frame_count = count_tar_pngs(root / "rgb.tar", "rgb")
+                if frame_count <= 0:
+                    continue
+                units.append(
+                    Unit(
+                        kind="ego",
+                        unit_id=f"{root.parent.name}.ego",
+                        root=root,
+                        expected_frames=frame_count,
+                        view="ego",
+                        scene_id=root.parent.name,
+                        phase_name="ego",
+                    )
+                )
+
+    mask_map_rows = read_csv_rows(repo_root / "manifest" / "mos_mask_object_map_v1.csv")
+    mask_map_by_phase: dict[tuple[str, str], list[dict[str, str]]] = defaultdict(list)
+    for row in mask_map_rows:
+        mask_map_by_phase[(row["scene_id"], row["phase"])].append(row)
+    context["mask_map_rows"] = mask_map_rows
+    context["mask_map_by_phase"] = mask_map_by_phase
+    ego_mask_map_path = repo_root / "manifest" / "mos_ego_mask_object_map_v1.csv"
+    ego_mask_map_rows = read_csv_rows(ego_mask_map_path) if ego_mask_map_path.exists() else []
+    ego_mask_map_by_scene: dict[str, list[dict[str, str]]] = defaultdict(list)
+    for row in ego_mask_map_rows:
+        ego_mask_map_by_scene[row["scene_id"]].append(row)
+    context["ego_mask_map_rows"] = ego_mask_map_rows
+    context["ego_mask_map_by_scene"] = ego_mask_map_by_scene
+    return units, context
+
+
+def check_global_manifests(repo_root: Path, units: list[Unit], context: dict[str, Any]) -> dict[str, Any]:
+    errors: list[str] = []
+    warnings: list[str] = []
+    summary: dict[str, Any] = {"errors": errors, "warnings": warnings}
+
+    current_path = repo_root / "manifest" / "current.json"
+    try:
+        current = load_json(current_path)
+        summary["current_json"] = {
+            "schema_version": current.get("schema_version"),
+            "label_versions": current.get("label_versions"),
+            "mos": current.get("mos"),
+            "sos": current.get("sos"),
+        }
+        expected_versions = {
+            "masks": LABEL_VERSION,
+            "masks_aux": LABEL_VERSION,
+            "sam2_meta": LABEL_VERSION,
+            "cam_pose": LABEL_VERSION,
+        }
+        if current.get("label_versions") != expected_versions:
+            errors.append(f"manifest/current.json label_versions != {expected_versions}")
+    except Exception as exc:
+        errors.append(f"could not read manifest/current.json: {exc}")
+
+    split_rows = context["split_rows"]
+    phase_ids = [row["scene_phase"] for row in split_rows]
+    difficulty_counts = Counter(row["difficulty"] for row in split_rows)
+    split_counts = Counter(row["split"] for row in split_rows)
+    summary["phase_splits"] = {
+        "rows": len(split_rows),
+        "unique_phase_ids": len(set(phase_ids)),
+        "split_counts": dict(split_counts),
+        "difficulty_counts": dict(difficulty_counts),
+    }
+    if len(split_rows) != 300 or len(set(phase_ids)) != 300:
+        errors.append("phase split CSVs should contain 300 unique scene phases")
+    for split in ("easy", "medium", "hard"):
+        if split_counts[split] != 100:
+            errors.append(f"splits/{split}.csv has {split_counts[split]} rows, expected 100")
+        mismatch = [row["scene_phase"] for row in split_rows if row["split"] == split and row["difficulty"] != split]
+        if mismatch:
+            errors.append(f"splits/{split}.csv difficulty mismatch examples: {mismatch[:5]}")
+
+    scene_splits_path = repo_root / "splits" / "scene_splits.json"
+    try:
+        scene_splits = load_json(scene_splits_path)
+        weights = scene_splits.get("provenance", {}).get("weights", {})
+        feature_names = scene_splits.get("provenance", {}).get("feature_names", [])
+        tiers = scene_splits.get("splits", {})
+        tier_ids = [scene_id for ids in tiers.values() for scene_id in ids]
+        summary["scene_splits_json"] = {
+            "n_scenes": scene_splits.get("n_scenes"),
+            "primary_method": scene_splits.get("primary_method"),
+            "scoring_version": scene_splits.get("scoring_version"),
+            "feature_count": len(feature_names),
+            "weight_count": len(weights),
+            "weight_sum": round(float(sum(weights.values())), 8) if weights else None,
+            "tier_counts": {key: len(value) for key, value in tiers.items()},
+        }
+        if len(feature_names) != 27 or len(weights) != 27:
+            errors.append("scene_splits.json should expose 27 ESD features and 27 weights")
+        if set(feature_names) != set(weights):
+            errors.append("scene_splits.json feature_names and weight keys differ")
+        if weights and not math.isclose(sum(weights.values()), 1.0, abs_tol=1e-6):
+            errors.append(f"scene_splits.json weights sum to {sum(weights.values())}, expected 1.0")
+        if len(tier_ids) != 100 or len(set(tier_ids)) != 100:
+            errors.append("scene_splits.json should contain 100 unique scene-tier IDs")
+    except Exception as exc:
+        errors.append(f"could not read splits/scene_splits.json: {exc}")
+
+    preview_csv = repo_root / "preview" / "mos_phase_preview.csv"
+    if not preview_csv.exists():
+        preview_csv = repo_root / "preview" / "data_studio_preview.csv"
+    try:
+        preview_rows = read_csv_rows(preview_csv)
+        preview_errors = 0
+        preview_sizes = Counter()
+        for row in preview_rows:
+            image_name = Path(row["image_preview_url"]).name
+            image_path = repo_root / "preview" / "image_examples" / "preview" / "images" / image_name
+            if not image_path.exists():
+                preview_errors += 1
+                continue
+            try:
+                with Image.open(image_path) as image:
+                    preview_sizes[str(image.size)] += 1
+            except Exception:
+                preview_errors += 1
+        summary["data_studio_preview"] = {
+            "rows": len(preview_rows),
+            "unique_scene_phases": len({row["scene_phase"] for row in preview_rows}),
+            "image_decode_errors": preview_errors,
+            "image_sizes": dict(preview_sizes),
+        }
+        if len(preview_rows) != 300:
+            errors.append(f"preview/data_studio_preview.csv has {len(preview_rows)} rows, expected 300")
+        if preview_errors:
+            errors.append(f"Data Studio preview image decode/missing errors: {preview_errors}")
+    except Exception as exc:
+        errors.append(f"could not read preview/data_studio_preview.csv: {exc}")
+
+    selected_rows = context["selected_sos_rows"]
+    global_ids = sorted(int(row["global_object_id"]) for row in selected_rows)
+    summary["selected_sos_objects"] = {
+        "rows": len(selected_rows),
+        "unique_object_ids": len({row["object_id"] for row in selected_rows}),
+        "global_id_min": global_ids[0] if global_ids else None,
+        "global_id_max": global_ids[-1] if global_ids else None,
+    }
+    if len(selected_rows) != 70 or global_ids != list(range(1, 71)):
+        errors.append("selected_sos_objects_v1.csv should contain global IDs 1..70")
+
+    try:
+        object_catalog = load_json(repo_root / "manifest" / "object_catalog_v1.json")
+        catalog_objects = object_catalog.get("objects", [])
+        summary["object_catalog"] = {"objects": len(catalog_objects)}
+        if len(catalog_objects) != 70:
+            errors.append("object_catalog_v1.json should contain 70 selected objects")
+    except Exception as exc:
+        errors.append(f"could not read object_catalog_v1.json: {exc}")
+
+    try:
+        meta_index = load_json(repo_root / "objects_meta" / "_index.json")
+        summary["objects_meta_index"] = {
+            "selected_object_count": meta_index.get("selected_object_count"),
+            "object_ids": len(meta_index.get("object_ids", [])),
+        }
+        if meta_index.get("selected_object_count") != 70:
+            errors.append("objects_meta/_index.json selected_object_count should be 70")
+    except Exception as exc:
+        errors.append(f"could not read objects_meta/_index.json: {exc}")
+
+    mask_rows = context["mask_map_rows"]
+    summary["mos_mask_object_map"] = {
+        "rows": len(mask_rows),
+        "unique_scene_phases": len({(row["scene_id"], row["phase"]) for row in mask_rows}),
+    }
+    if len(mask_rows) != 2100:
+        errors.append(f"mos_mask_object_map_v1.csv has {len(mask_rows)} rows, expected 2100")
+
+    ego_frame_rows = context.get("ego_frame_rows", [])
+    ego_mask_rows = context.get("ego_mask_map_rows", [])
+    if ego_frame_rows:
+        summary["ego_frames"] = {
+            "rows": len(ego_frame_rows),
+            "unique_scenes": len({row["scene_id"] for row in ego_frame_rows}),
+        }
+    if ego_mask_rows:
+        summary["mos_ego_mask_object_map"] = {
+            "rows": len(ego_mask_rows),
+            "unique_scenes": len({row["scene_id"] for row in ego_mask_rows}),
+        }
+
+    ego_preview_csv = repo_root / "preview" / "ego_preview.csv"
+    if ego_preview_csv.exists():
+        try:
+            ego_preview_rows = read_csv_rows(ego_preview_csv)
+            summary["ego_preview"] = {
+                "rows": len(ego_preview_rows),
+                "unique_scenes": len({row["scene_id"] for row in ego_preview_rows}),
+            }
+        except Exception as exc:
+            errors.append(f"could not read preview/ego_preview.csv: {exc}")
+
+    for path in [repo_root / "assets" / "rpx_teaser.png", repo_root / "assets" / "rpx-jumbotron.webm"]:
+        if not path.exists() or path.stat().st_size == 0:
+            errors.append(f"missing or empty asset: {path}")
+    try:
+        with Image.open(repo_root / "assets" / "rpx_teaser.png") as image:
+            summary["teaser_image"] = {"size": list(image.size), "mode": image.mode}
+    except Exception as exc:
+        errors.append(f"could not decode assets/rpx_teaser.png: {exc}")
+
+    parquet_summary = check_parquet_manifests(repo_root, units, warnings, errors)
+    summary["parquet"] = parquet_summary
+    summary["status"] = "error" if errors else ("warning" if warnings else "ok")
+    return summary
+
+
+def check_parquet_manifests(
+    repo_root: Path,
+    units: list[Unit],
+    warnings: list[str],
+    errors: list[str],
+) -> dict[str, Any]:
+    summary: dict[str, Any] = {}
+    try:
+        import pyarrow.parquet as pq
+    except Exception as exc:
+        warnings.append(f"pyarrow unavailable; parquet row-count checks skipped: {exc}")
+        return {"skipped": True, "reason": str(exc)}
+
+    expected_rows = {
+        "manifest/frames_v1.parquet": 110000,
+        "preview/mos_phase_preview.parquet": 300,
+        "manifest/selected_sos_objects_v1.parquet": 70,
+        "splits/easy.parquet": 100,
+        "splits/medium.parquet": 100,
+        "splits/hard.parquet": 100,
+    }
+    if not (repo_root / "preview" / "mos_phase_preview.parquet").exists():
+        expected_rows["preview/media_preview.parquet"] = 300
+        expected_rows.pop("preview/mos_phase_preview.parquet", None)
+    ego_frames_csv = repo_root / "manifest" / "ego_frames_v1.csv"
+    ego_frame_rows = read_csv_rows(ego_frames_csv) if ego_frames_csv.exists() else []
+    if ego_frame_rows:
+        expected_rows["manifest/ego_frames_v1.parquet"] = len(ego_frame_rows)
+        expected_rows["manifest/frames_v2.parquet"] = 110000 + len(ego_frame_rows)
+    ego_mask_csv = repo_root / "manifest" / "mos_ego_mask_object_map_v1.csv"
+    if ego_mask_csv.exists():
+        expected_rows["manifest/mos_ego_mask_object_map_v1.parquet"] = len(read_csv_rows(ego_mask_csv))
+    ego_preview_csv = repo_root / "preview" / "ego_preview.csv"
+    if ego_preview_csv.exists():
+        expected_rows["preview/ego_preview.parquet"] = len(read_csv_rows(ego_preview_csv))
+    for rel, expected in expected_rows.items():
+        path = repo_root / rel
+        if not path.exists():
+            errors.append(f"missing parquet: {rel}")
+            continue
+        try:
+            pf = pq.ParquetFile(path)
+            row_count = pf.metadata.num_rows
+            summary[rel] = {"rows": row_count, "columns": pf.schema_arrow.names}
+            if row_count != expected:
+                errors.append(f"{rel} row count {row_count}, expected {expected}")
+        except Exception as exc:
+            errors.append(f"could not read {rel}: {exc}")
+
+    frames_path = repo_root / "manifest" / "frames_v1.parquet"
+    if frames_path.exists():
+        try:
+            table = pq.read_table(
+                frames_path,
+                columns=[
+                    "scene_id",
+                    "scene_type",
+                    "phase",
+                    "has_rgb",
+                    "has_depth",
+                    "has_fisheye",
+                    "has_cam_pose",
+                    "has_masks",
+                    "has_masks_aux",
+                    "has_sam2_meta",
+                ],
+            )
+            data = {name: table[name].to_pylist() for name in table.column_names}
+            unit_counts: Counter[tuple[str, str, int]] = Counter()
+            false_counts: Counter[str] = Counter()
+            for index in range(table.num_rows):
+                unit_counts[(data["scene_type"][index], data["scene_id"][index], int(data["phase"][index]))] += 1
+                for column in (
+                    "has_rgb",
+                    "has_depth",
+                    "has_fisheye",
+                    "has_cam_pose",
+                    "has_masks",
+                    "has_masks_aux",
+                    "has_sam2_meta",
+                ):
+                    if not data[column][index]:
+                        false_counts[column] += 1
+
+            frame_count_mismatches = []
+            for unit in units:
+                if unit.kind == "mos":
+                    key = ("multi_object", unit.scene_id or "", unit.phase_index)
+                elif unit.kind == "sos":
+                    key = ("single_object", unit.object_id or "", 0)
+                else:
+                    continue
+                count = unit_counts.get(key, 0)
+                if count != unit.expected_frames:
+                    frame_count_mismatches.append(
+                        {"unit_id": unit.unit_id, "manifest_frames": count, "expected": unit.expected_frames}
+                    )
+            summary["manifest/frames_v1.parquet"]["unit_frame_count_mismatches"] = frame_count_mismatches[:20]
+            summary["manifest/frames_v1.parquet"]["false_modality_flags"] = dict(false_counts)
+            if frame_count_mismatches:
+                errors.append(f"frames_v1.parquet unit frame mismatches: {len(frame_count_mismatches)}")
+            if false_counts:
+                errors.append(f"frames_v1.parquet has false modality flags: {dict(false_counts)}")
+        except Exception as exc:
+            errors.append(f"could not validate frames_v1.parquet unit rows: {exc}")
+    return summary
+
+
+def check_image_stream(
+    shard_path: Path,
+    names: list[str],
+    prefix: str,
+    expected_count: int,
+    expected_size: tuple[int, int],
+    policy: str,
+    errors: list[str],
+    warnings: list[str],
+    label: str,
+    require_nonzero: bool = False,
+) -> tuple[dict[str, Any], dict[str, Image.Image]]:
+    stream_names = names_with_prefix(names, prefix, ".png")
+    sequence = numbered_sequence_summary(stream_names, prefix, ".png", expected_count, errors, label)
+    sample_names = pick_names(stream_names, policy)
+    sample_stats, sample_images = decode_images(
+        shard_path,
+        sample_names,
+        errors,
+        warnings,
+        label,
+        expected_size=expected_size,
+        require_nonzero=require_nonzero,
+    )
+    return {
+        "prefix": prefix,
+        **sequence,
+        "sampled": sample_stats,
+    }, sample_images
+
+
+def check_masks_aux(
+    shard_path: Path,
+    names: list[str],
+    unit: Unit,
+    policy: str,
+    errors: list[str],
+    warnings: list[str],
+) -> dict[str, Any]:
+    group_counts: Counter[str] = Counter()
+    for name in names:
+        parts = name.split("/")
+        key = "/".join(parts[:2]) if len(parts) >= 2 else name
+        group_counts[key] += 1
+
+    summary: dict[str, Any] = {"file_count": len(names), "group_counts": dict(sorted(group_counts.items()))}
+    expected = unit.expected_frames
+    frame_aligned_groups = ["sam2/contour_gt_masks", "sam2/palette", "sam2/rgb_and_mask"]
+    if unit.kind in {"mos", "ego"}:
+        frame_aligned_groups.append("sam2/masks_contour_with_hidden")
+
+    decoded_groups: dict[str, Any] = {}
+    for group in frame_aligned_groups:
+        group_names = names_with_prefix(names, group, ".png")
+        sequence = numbered_sequence_summary(
+            group_names,
+            group,
+            ".png",
+            expected,
+            errors,
+            f"{unit.unit_id} masks_aux:{group}",
+        )
+        sample_stats, _ = decode_images(
+            shard_path,
+            pick_names(group_names, policy),
+            errors,
+            warnings,
+            f"{unit.unit_id} masks_aux:{group}",
+            expected_size=RGBD_SIZE,
+        )
+        decoded_groups[group] = {**sequence, "sampled": sample_stats}
+
+    bbox_names = names_with_prefix(names, "sam2/bbox_overlay", ".png")
+    decoded_bbox, _ = decode_images(
+        shard_path,
+        bbox_names[:3],
+        errors,
+        warnings,
+        f"{unit.unit_id} masks_aux:sam2/bbox_overlay",
+        expected_size=RGBD_SIZE,
+    )
+    decoded_groups["sam2/bbox_overlay"] = {"count": len(bbox_names), "sampled": decoded_bbox}
+
+    dino_names = [name for name in names if name.startswith("sam2/dino_output/")]
+    dino_png_names = [name for name in dino_names if name.lower().endswith(".png")]
+    dino_json_names = [name for name in dino_names if name.lower().endswith(".json")]
+    if unit.kind in {"mos", "ego"}:
+        if not dino_names:
+            warnings.append(f"{unit.unit_id} masks_aux:sam2/dino_output is missing")
+        decoded_dino, _ = decode_images(
+            shard_path,
+            dino_png_names[:3],
+            errors,
+            warnings,
+            f"{unit.unit_id} masks_aux:sam2/dino_output",
+        )
+        decoded_groups["sam2/dino_output"] = {
+            "count": len(dino_names),
+            "png_count": len(dino_png_names),
+            "json_count": len(dino_json_names),
+            "sampled": decoded_dino,
+        }
+    elif dino_names:
+        decoded_dino, _ = decode_images(
+            shard_path,
+            dino_png_names[:3],
+            errors,
+            warnings,
+            f"{unit.unit_id} masks_aux:sam2/dino_output",
+        )
+        decoded_groups["sam2/dino_output"] = {
+            "count": len(dino_names),
+            "png_count": len(dino_png_names),
+            "json_count": len(dino_json_names),
+            "sampled": decoded_dino,
+        }
+
+    summary["decoded_groups"] = decoded_groups
+    return summary
+
+
+def check_sam2_meta(
+    shard_path: Path,
+    names: list[str],
+    unit: Unit,
+    context: dict[str, Any],
+    errors: list[str],
+    warnings: list[str],
+) -> dict[str, Any]:
+    summary: dict[str, Any] = {"file_count": len(names), "files": names}
+    required = (
+        ["sam2/mask_to_object.json"]
+        if unit.kind in {"mos", "ego"}
+        else ["sam2/iter1_faulty.txt", "sam2/usr_bbox_prompts.npy", "sam2/verify_page.txt"]
+    )
+    missing = [name for name in required if name not in names]
+    if missing:
+        errors.append(f"{unit.unit_id} sam2_meta missing files: {missing}")
+
+    try:
+        with tarfile.open(shard_path) as tar:
+            if unit.kind in {"mos", "ego"} and "sam2/mask_to_object.json" in names:
+                mask_to_object = read_tar_json(tar, "sam2/mask_to_object.json")
+                summary["mask_to_object_count"] = len(mask_to_object)
+                if unit.kind == "mos":
+                    expected_rows = context["mask_map_by_phase"].get((unit.scene_id, f"phase{unit.phase_index}"), [])
+                else:
+                    expected_rows = context.get("ego_mask_map_by_scene", {}).get(unit.scene_id, [])
+                expected_ids = {int(row["local_mask_id"]) for row in expected_rows}
+                actual_ids = {int(key) for key in mask_to_object.keys()}
+                if len(expected_rows) != len(mask_to_object):
+                    errors.append(
+                        f"{unit.unit_id} mask map row count {len(expected_rows)}, mask_to_object count {len(mask_to_object)}"
+                    )
+                if expected_ids != actual_ids:
+                    errors.append(f"{unit.unit_id} local mask IDs differ between manifest and sam2_meta")
+            if unit.kind in {"mos", "ego"}:
+                line_counts = {}
+                for name in names:
+                    if name.endswith(".txt") and name in names:
+                        text = read_tar_text(tar, name)
+                        line_counts[name] = len([line for line in text.splitlines() if line.strip()])
+                summary["text_line_counts"] = line_counts
+            else:
+                if "sam2/usr_bbox_prompts.npy" in names:
+                    extracted = tar.extractfile("sam2/usr_bbox_prompts.npy")
+                    if extracted is None:
+                        errors.append(f"{unit.unit_id} could not extract usr_bbox_prompts.npy")
+                    else:
+                        bbox = np.load(io.BytesIO(extracted.read()), allow_pickle=True)
+                        summary["usr_bbox_prompts"] = {
+                            "shape": list(bbox.shape),
+                            "dtype": str(bbox.dtype),
+                            "finite": bool(np.isfinite(bbox).all()) if np.issubdtype(bbox.dtype, np.number) else None,
+                        }
+                        if bbox.ndim != 2 or bbox.shape[-1] != 4:
+                            errors.append(f"{unit.unit_id} usr_bbox_prompts.npy shape {bbox.shape}, expected Nx4")
+                if "sam2/verify_page.txt" in names:
+                    summary["verify_page"] = read_tar_text(tar, "sam2/verify_page.txt").strip()
+    except Exception as exc:
+        errors.append(f"{unit.unit_id} sam2_meta could not be parsed: {exc}")
+    return summary
+
+
+def check_questionnaire(repo_root: Path, unit: Unit, errors: list[str]) -> dict[str, Any]:
+    if unit.kind != "sos" or unit.object_id is None:
+        return {}
+    meta_path = repo_root / "objects_meta" / unit.object_id / "metadata.json"
+    questionnaire_path = repo_root / "objects_meta" / unit.object_id / "questionnaire.json"
+    summary: dict[str, Any] = {}
+    for label, path in [("metadata", meta_path), ("questionnaire", questionnaire_path)]:
+        if not path.exists():
+            errors.append(f"{unit.unit_id} missing {label}: {path}")
+            continue
+        try:
+            data = load_json(path)
+            summary[label] = {
+                "path": as_rel(path, repo_root),
+                "object_id": data.get("object_id"),
+                "global_object_id": data.get("global_object_id"),
+            }
+            if data.get("object_id") != unit.object_id:
+                errors.append(f"{unit.unit_id} {label} object_id mismatch")
+            if data.get("global_object_id") != unit.global_object_id:
+                errors.append(f"{unit.unit_id} {label} global_object_id mismatch")
+            if label == "questionnaire":
+                questions = data.get("questions", {})
+                summary[label]["question_count"] = len(questions)
+                if not questions:
+                    errors.append(f"{unit.unit_id} questionnaire has no questions")
+        except Exception as exc:
+            errors.append(f"{unit.unit_id} could not parse {label}: {exc}")
+    return summary
+
+
+def check_unit(
+    repo_root: Path,
+    unit: Unit,
+    context: dict[str, Any],
+    policy: str,
+) -> tuple[dict[str, Any], dict[str, Image.Image]]:
+    errors: list[str] = []
+    warnings: list[str] = []
+    root = unit.root
+    result: dict[str, Any] = {
+        "unit_id": unit.unit_id,
+        "kind": unit.kind,
+        "view": unit.view,
+        "root": as_rel(root, repo_root),
+        "expected_frames": unit.expected_frames,
+        "errors": errors,
+        "warnings": warnings,
+        "shards": {},
+    }
+    viz: dict[str, Image.Image] = {}
+    if not root.exists():
+        errors.append(f"missing unit directory: {root}")
+
+    shard_paths = {
+        "rgb": root / "rgb.tar",
+        "depth": root / "depth.tar",
+        "fisheye": root / "fisheye.tar",
+        "masks": root / "labels" / "masks" / f"{LABEL_VERSION}.tar",
+        "masks_aux": root / "labels" / "masks_aux" / f"{LABEL_VERSION}.tar",
+        "sam2_meta": root / "labels" / "sam2_meta" / f"{LABEL_VERSION}.tar",
+        "cam_pose": root / "labels" / "cam_pose" / f"{LABEL_VERSION}.tar",
+    }
+
+    tar_names: dict[str, list[str]] = {}
+    for key, path in shard_paths.items():
+        names = tar_file_names(path, errors)
+        tar_names[key] = names
+        result["shards"][key] = {
+            "path": as_rel(path, repo_root),
+            "file_count": len(names),
+        }
+
+    rgb_summary, rgb_images = check_image_stream(
+        shard_paths["rgb"],
+        tar_names["rgb"],
+        "rgb",
+        unit.expected_frames,
+        RGBD_SIZE,
+        policy,
+        errors,
+        warnings,
+        f"{unit.unit_id} rgb",
+    )
+    result["shards"]["rgb"].update(rgb_summary)
+    if rgb_images:
+        viz["rgb"] = rgb_images[pick_names(names_with_prefix(tar_names["rgb"], "rgb", ".png"), "middle")[0]]
+
+    depth_summary, depth_images = check_image_stream(
+        shard_paths["depth"],
+        tar_names["depth"],
+        "depth",
+        unit.expected_frames,
+        RGBD_SIZE,
+        policy,
+        errors,
+        warnings,
+        f"{unit.unit_id} depth",
+        require_nonzero=True,
+    )
+    result["shards"]["depth"].update(depth_summary)
+    if depth_images:
+        viz["depth"] = depth_images[pick_names(names_with_prefix(tar_names["depth"], "depth", ".png"), "middle")[0]]
+
+    mask_summary, mask_images = check_image_stream(
+        shard_paths["masks"],
+        tar_names["masks"],
+        "sam2/masks",
+        unit.expected_frames,
+        RGBD_SIZE,
+        policy,
+        errors,
+        warnings,
+        f"{unit.unit_id} masks",
+        require_nonzero=True,
+    )
+    result["shards"]["masks"].update(mask_summary)
+    mask_sample_stats = result["shards"]["masks"].get("sampled", [])
+    for item in mask_sample_stats:
+        # Mask PNGs are integer maps; count nonzero IDs on sampled frames.
+        name = item["name"]
+        image = mask_images.get(name)
+        if image is not None:
+            values = np.unique(np.asarray(image))
+            item["nonzero_instance_ids"] = int(np.count_nonzero(values))
+    if mask_images:
+        viz["mask"] = mask_images[pick_names(names_with_prefix(tar_names["masks"], "sam2/masks", ".png"), "middle")[0]]
+
+    fisheye_left_summary, fisheye_left_images = check_image_stream(
+        shard_paths["fisheye"],
+        tar_names["fisheye"],
+        "fisheye/left",
+        unit.expected_frames,
+        FISHEYE_SIZE,
+        policy,
+        errors,
+        warnings,
+        f"{unit.unit_id} fisheye_left",
+    )
+    fisheye_right_summary, fisheye_right_images = check_image_stream(
+        shard_paths["fisheye"],
+        tar_names["fisheye"],
+        "fisheye/right",
+        unit.expected_frames,
+        FISHEYE_SIZE,
+        policy,
+        errors,
+        warnings,
+        f"{unit.unit_id} fisheye_right",
+    )
+    result["shards"]["fisheye"].update(
+        {
+            "left": fisheye_left_summary,
+            "right": fisheye_right_summary,
+        }
+    )
+    left_names = names_with_prefix(tar_names["fisheye"], "fisheye/left", ".png")
+    right_names = names_with_prefix(tar_names["fisheye"], "fisheye/right", ".png")
+    if fisheye_left_images and left_names:
+        viz["fisheye_left"] = fisheye_left_images[pick_names(left_names, "middle")[0]]
+    if fisheye_right_images and right_names:
+        viz["fisheye_right"] = fisheye_right_images[pick_names(right_names, "middle")[0]]
+
+    cam_pose_names = names_with_prefix(tar_names["cam_pose"], "cam_pose", ".npz")
+    cam_sequence = numbered_sequence_summary(
+        cam_pose_names,
+        "cam_pose",
+        ".npz",
+        unit.expected_frames,
+        errors,
+        f"{unit.unit_id} cam_pose",
+    )
+    result["shards"]["cam_pose"].update(
+        {
+            **cam_sequence,
+            "sampled": decode_npz_samples(
+                shard_paths["cam_pose"],
+                pick_names(cam_pose_names, policy),
+                errors,
+                warnings,
+                f"{unit.unit_id} cam_pose",
+            ),
+        }
+    )
+
+    result["shards"]["masks_aux"].update(
+        check_masks_aux(shard_paths["masks_aux"], tar_names["masks_aux"], unit, policy, errors, warnings)
+    )
+    result["shards"]["sam2_meta"].update(
+        check_sam2_meta(shard_paths["sam2_meta"], tar_names["sam2_meta"], unit, context, errors, warnings)
+    )
+    if unit.kind == "sos":
+        result["object_metadata"] = check_questionnaire(repo_root, unit, errors)
+
+    frame_counts = {
+        "rgb": result["shards"]["rgb"].get("count"),
+        "depth": result["shards"]["depth"].get("count"),
+        "masks": result["shards"]["masks"].get("count"),
+        "cam_pose": result["shards"]["cam_pose"].get("count"),
+        "fisheye_left": result["shards"]["fisheye"].get("left", {}).get("count"),
+        "fisheye_right": result["shards"]["fisheye"].get("right", {}).get("count"),
+    }
+    if any(count != unit.expected_frames for count in frame_counts.values()):
+        errors.append(f"{unit.unit_id} frame count alignment failed: {frame_counts}")
+    result["frame_counts"] = frame_counts
+    result["status"] = "error" if errors else ("warning" if warnings else "ok")
+    return result, viz
+
+
+def write_markdown_summary(summary: dict[str, Any], out_path: Path) -> None:
+    lines = [
+        "# RPX All-Modality QA Summary",
+        "",
+        f"- Status: `{summary['status']}`",
+        f"- Units checked: `{summary['units_checked']}`",
+        f"- MOS units: `{summary['unit_counts'].get('mos', 0)}`",
+        f"- Ego units: `{summary['unit_counts'].get('ego', 0)}`",
+        f"- SOS units: `{summary['unit_counts'].get('sos', 0)}`",
+        f"- Errors: `{summary['error_count']}`",
+        f"- Warnings: `{summary['warning_count']}`",
+        f"- Rerun recording: `{summary.get('rrd_path')}`",
+        f"- Contact sheet: `{summary.get('contact_sheet_path')}`",
+        "",
+        "## Coverage",
+        "",
+        "- RGB, depth, masks, fisheye left/right, camera poses, masks_aux, and SAM2 metadata shards.",
+        "- MOS split tables, scene ESD weights/tier JSON, Data Studio preview images, ego previews, and mask-object manifest joins.",
+        "- SOS selected-object table, object catalog, object metadata, and questionnaires.",
+        "- Parquet row counts and frame-manifest unit counts when pyarrow is available.",
+        "",
+    ]
+    if summary["errors_by_unit"]:
+        lines.extend(["## Error Examples", ""])
+        for item in summary["errors_by_unit"][:20]:
+            lines.append(f"- `{item['unit_id']}`: {item['errors'][0]}")
+        lines.append("")
+    if summary["warnings_by_unit"]:
+        lines.extend(["## Warning Examples", ""])
+        for item in summary["warnings_by_unit"][:20]:
+            lines.append(f"- `{item['unit_id']}`: {item['warnings'][0]}")
+        lines.append("")
+    out_path.write_text("\n".join(lines), encoding="utf-8")
+
+
+def main() -> None:
+    args = parse_args()
+    repo_root = args.repo_root.resolve()
+    out_dir = repo_path(repo_root, args.out_dir)
+    out_dir.mkdir(parents=True, exist_ok=True)
+    kinds = {item.strip() for item in args.kinds.split(",") if item.strip()}
+    invalid_kinds = kinds - {"mos", "sos", "ego"}
+    if invalid_kinds:
+        raise SystemExit(f"invalid --kinds values: {sorted(invalid_kinds)}")
+
+    units, context = build_units(repo_root, kinds)
+    if args.limit is not None:
+        units = units[: args.limit]
+
+    report_path = out_dir / "all_modalities_report.json"
+    markdown_path = out_dir / "all_modalities_summary.md"
+    contact_sheet_path = out_dir / "all_modalities_contact_sheet.jpg"
+    rrd_path = out_dir / "rpx_all_modalities_visual_check.rrd"
+
+    global_summary = check_global_manifests(repo_root, units, context)
+    rr = None if args.skip_rerun else open_rerun(rrd_path)
+    tiles: list[Image.Image] = []
+    results: list[dict[str, Any]] = []
+    for index, unit in enumerate(tqdm(units, desc="checking RPX units")):
+        result, viz = check_unit(repo_root, unit, context, args.sample_policy)
+        results.append(result)
+        if not args.skip_contact_sheet:
+            rendered_viz = dict(viz)
+            if "depth" in rendered_viz:
+                rendered_viz["depth"] = depth_to_uint8(rendered_viz["depth"])
+            if "mask" in rendered_viz:
+                rendered_viz["mask"] = colorize_mask(rendered_viz["mask"])
+            tiles.append(make_unit_tile(unit, result, rendered_viz, args.contact_thumb_width))
+        if rr is not None:
+            log_to_rerun(rr, index, unit, result, viz, args.image_width)
+
+    if tiles:
+        make_contact_sheet(tiles, contact_sheet_path, args.contact_cols)
+
+    unit_counts = Counter(result["kind"] for result in results)
+    errors_by_unit = [
+        {"unit_id": result["unit_id"], "errors": result["errors"]}
+        for result in results
+        if result["errors"]
+    ]
+    warnings_by_unit = [
+        {"unit_id": result["unit_id"], "warnings": result["warnings"]}
+        for result in results
+        if result["warnings"]
+    ]
+    error_count = sum(len(result["errors"]) for result in results) + len(global_summary["errors"])
+    warning_count = sum(len(result["warnings"]) for result in results) + len(global_summary["warnings"])
+    status = "error" if error_count else ("warning" if warning_count else "ok")
+    summary = {
+        "repo_root": str(repo_root),
+        "sample_policy": args.sample_policy,
+        "status": status,
+        "units_checked": len(results),
+        "unit_counts": dict(unit_counts),
+        "error_count": error_count,
+        "warning_count": warning_count,
+        "global": global_summary,
+        "errors_by_unit": errors_by_unit,
+        "warnings_by_unit": warnings_by_unit,
+        "rrd_path": str(rrd_path) if rr is not None else None,
+        "contact_sheet_path": str(contact_sheet_path) if tiles else None,
+        "results": results,
+    }
+    report_path.write_text(json.dumps(summary, indent=2), encoding="utf-8")
+    write_markdown_summary(summary, markdown_path)
+
+    print(f"status: {status}")
+    print(f"units checked: {len(results)}")
+    print(f"errors: {error_count}")
+    print(f"warnings: {warning_count}")
+    print(f"wrote {report_path}")
+    print(f"wrote {markdown_path}")
+    if tiles:
+        print(f"wrote {contact_sheet_path}")
+    if rr is not None:
+        print(f"wrote {rrd_path}")
+    if error_count or (warning_count and args.fail_on_warnings):
+        raise SystemExit(1)
+
+
+if __name__ == "__main__":
+    main()

rerun_check/download_ego.py

@@ -0,0 +1,70 @@
+#!/usr/bin/env python3
+"""Download RPX egocentric MOS shards through the Hugging Face Hub API."""
+
+from __future__ import annotations
+
+import argparse
+from pathlib import Path
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--repo-id", default="IRVLUTD/RPX")
+    parser.add_argument(
+        "--scenes",
+        nargs="+",
+        default=None,
+        help="Scene IDs to download, e.g. scene001 scene002. Default: all ego scene shards.",
+    )
+    parser.add_argument("--local-dir", type=Path, default=Path("rpx_ego_download"))
+    parser.add_argument(
+        "--include-preview",
+        action="store_true",
+        help="Also download ego preview/Data Studio files.",
+    )
+    return parser.parse_args()
+
+
+def main() -> None:
+    args = parse_args()
+
+    from huggingface_hub import snapshot_download
+
+    scene_patterns = (
+        [f"scenes/{scene_id}/ego/**" for scene_id in args.scenes]
+        if args.scenes
+        else ["scenes/*/ego/**"]
+    )
+    allow_patterns = [
+        "README.md",
+        "manifest/current.json",
+        "manifest/ego_frames_v1.csv",
+        "manifest/ego_frames_v1.parquet",
+        "manifest/frames_v2.parquet",
+        "manifest/mos_ego_mask_object_map_v1.csv",
+        "manifest/mos_ego_mask_object_map_v1.parquet",
+        *scene_patterns,
+    ]
+    if args.include_preview:
+        allow_patterns.extend(
+            [
+                "preview/ego_preview.csv",
+                "preview/ego_preview.parquet",
+                "preview/image_examples/ego_preview/**",
+            ]
+        )
+
+    local_dir = snapshot_download(
+        repo_id=args.repo_id,
+        repo_type="dataset",
+        local_dir=args.local_dir,
+        allow_patterns=allow_patterns,
+    )
+    print(local_dir)
+    print("downloaded ego patterns:")
+    for pattern in allow_patterns:
+        print(f"  {pattern}")
+
+
+if __name__ == "__main__":
+    main()

rerun_check/download_rgbd_segmentation.py

@@ -0,0 +1,40 @@
+#!/usr/bin/env python3
+"""Download RPX RGB-D segmentation shards through the published Quick Start API."""
+
+from __future__ import annotations
+
+import argparse
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(
+        description="Download RPX rgbd_segmentation shards for all difficulty splits."
+    )
+    parser.add_argument("--repo-id", default="IRVLUTD/RPX")
+    parser.add_argument(
+        "--splits",
+        nargs="+",
+        default=["easy", "medium", "hard"],
+        choices=["easy", "medium", "hard"],
+    )
+    return parser.parse_args()
+
+
+def main() -> None:
+    args = parse_args()
+
+    from rpx_benchmark.dataset_hub import download_for_task
+
+    for split in args.splits:
+        result = download_for_task(
+            task="rgbd_segmentation",
+            split=split,
+            repo_id=args.repo_id,
+        )
+        print(f"{split}: {result.local_dir}")
+        print(f"  matched scenes: {len(result.matched_scenes)}")
+
+
+if __name__ == "__main__":
+    main()
+

rerun_check/make_rerun_check.py

@@ -0,0 +1,379 @@
+#!/usr/bin/env python3
+"""Build a Rerun visual QA recording for RPX RGB/depth/mask samples.
+
+The default run checks one representative middle frame for every MOS
+scene-phase listed in preview/data_studio_preview.csv: 100 scenes x 3 phases.
+It writes:
+
+- out/rpx_visual_check.rrd: Rerun recording for interactive inspection.
+- out/rpx_visual_contact_sheet.jpg: overview grid using the preview composites.
+- out/visual_check_report.json: structural and per-sample QA summary.
+"""
+
+from __future__ import annotations
+
+import argparse
+import csv
+import json
+import tarfile
+from dataclasses import asdict, dataclass
+from pathlib import Path
+from typing import Iterable
+
+import numpy as np
+from PIL import Image, ImageDraw, ImageFont
+from tqdm import tqdm
+
+
+@dataclass
+class SampleResult:
+    index: int
+    scene_phase: str
+    scene_id: str
+    phase_index: int
+    difficulty: str
+    rgb_path: str
+    depth_path: str
+    mask_path: str
+    rgb_frames: int
+    depth_frames: int
+    mask_frames: int
+    frame_index: int
+    rgb_size: tuple[int, int]
+    depth_size: tuple[int, int]
+    mask_size: tuple[int, int]
+    depth_min: int
+    depth_max: int
+    depth_nonzero_fraction: float
+    mask_instance_count: int
+    mask_nonzero_fraction: float
+    status: str
+    notes: list[str]
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument(
+        "--repo-root",
+        type=Path,
+        default=Path(__file__).resolve().parents[1],
+        help="RPX dataset repository root or downloaded local_dir.",
+    )
+    parser.add_argument(
+        "--rows-csv",
+        type=Path,
+        default=Path("preview/data_studio_preview.csv"),
+        help="CSV containing scene_phase and source shard paths.",
+    )
+    parser.add_argument("--out-dir", type=Path, default=Path("rerun_check/out"))
+    parser.add_argument("--limit", type=int, default=None)
+    parser.add_argument(
+        "--one-per-scene",
+        action="store_true",
+        help="Sample only the first listed phase for each scene instead of all 300 scene-phases.",
+    )
+    parser.add_argument(
+        "--frame-index",
+        type=int,
+        default=None,
+        help="Frame index to inspect. Default: middle frame of each RGB tar.",
+    )
+    parser.add_argument("--image-width", type=int, default=480)
+    parser.add_argument("--contact-thumb-width", type=int, default=220)
+    parser.add_argument("--skip-rerun", action="store_true")
+    return parser.parse_args()
+
+
+def repo_path(repo_root: Path, value: str | Path) -> Path:
+    path = Path(value)
+    return path if path.is_absolute() else repo_root / path
+
+
+def load_rows(csv_path: Path, *, one_per_scene: bool, limit: int | None) -> list[dict[str, str]]:
+    rows: list[dict[str, str]] = []
+    seen_scenes: set[str] = set()
+    with csv_path.open(newline="") as handle:
+        for row in csv.DictReader(handle):
+            if one_per_scene and row["scene_id"] in seen_scenes:
+                continue
+            seen_scenes.add(row["scene_id"])
+            rows.append(row)
+            if limit is not None and len(rows) >= limit:
+                break
+    return rows
+
+
+def tar_png_names(path: Path) -> list[str]:
+    with tarfile.open(path) as tar:
+        return sorted(
+            member.name
+            for member in tar.getmembers()
+            if member.isfile() and member.name.lower().endswith(".png")
+        )
+
+
+def read_png_from_tar(path: Path, index: int) -> tuple[Image.Image, int, str]:
+    with tarfile.open(path) as tar:
+        names = sorted(
+            member.name
+            for member in tar.getmembers()
+            if member.isfile() and member.name.lower().endswith(".png")
+        )
+        if not names:
+            raise ValueError(f"no PNG frames in {path}")
+        safe_index = min(max(index, 0), len(names) - 1)
+        extracted = tar.extractfile(names[safe_index])
+        if extracted is None:
+            raise ValueError(f"could not read {names[safe_index]} from {path}")
+        image = Image.open(extracted).copy()
+    return image, len(names), names[safe_index]
+
+
+def resize_width(image: Image.Image, width: int, *, nearest: bool = False) -> Image.Image:
+    if image.width <= width:
+        return image.copy()
+    height = max(1, round(image.height * width / image.width))
+    method = Image.Resampling.NEAREST if nearest else Image.Resampling.LANCZOS
+    return image.resize((width, height), method)
+
+
+def depth_to_uint8(depth: Image.Image) -> Image.Image:
+    arr = np.asarray(depth).astype(np.float32)
+    valid = arr > 0
+    if not valid.any():
+        return Image.fromarray(np.zeros(arr.shape, dtype=np.uint8), mode="L")
+    lo, hi = np.percentile(arr[valid], [2, 98])
+    if hi <= lo:
+        hi = lo + 1
+    norm = np.clip((arr - lo) / (hi - lo), 0, 1)
+    norm[~valid] = 0
+    return Image.fromarray((norm * 255).astype(np.uint8), mode="L")
+
+
+def colorize_mask(mask: Image.Image) -> Image.Image:
+    arr = np.asarray(mask)
+    rgb = np.zeros((*arr.shape, 3), dtype=np.uint8)
+    palette = np.asarray(
+        [
+            [244, 103, 83],
+            [79, 210, 154],
+            [255, 184, 77],
+            [180, 101, 255],
+            [64, 182, 255],
+            [255, 99, 179],
+            [141, 224, 83],
+            [255, 226, 102],
+            [102, 232, 232],
+            [231, 91, 116],
+            [161, 135, 255],
+            [76, 217, 100],
+        ],
+        dtype=np.uint8,
+    )
+    for value in np.unique(arr):
+        if value == 0:
+            continue
+        rgb[arr == value] = palette[(int(value) - 1) % len(palette)]
+    return Image.fromarray(rgb, mode="RGB")
+
+
+def sample_status(
+    row: dict[str, str],
+    rgb: Image.Image,
+    depth: Image.Image,
+    mask: Image.Image,
+    rgb_frames: int,
+    depth_frames: int,
+    mask_frames: int,
+    frame_index: int,
+    rgb_path: Path,
+    depth_path: Path,
+    mask_path: Path,
+) -> SampleResult:
+    notes: list[str] = []
+    if not (rgb.size == depth.size == mask.size):
+        notes.append(f"size mismatch: rgb={rgb.size} depth={depth.size} mask={mask.size}")
+    if not (rgb_frames == depth_frames == mask_frames):
+        notes.append(
+            f"frame count mismatch: rgb={rgb_frames} depth={depth_frames} mask={mask_frames}"
+        )
+
+    depth_arr = np.asarray(depth)
+    mask_arr = np.asarray(mask)
+    depth_nonzero_fraction = float(np.count_nonzero(depth_arr) / depth_arr.size)
+    mask_nonzero_fraction = float(np.count_nonzero(mask_arr) / mask_arr.size)
+    mask_values = np.unique(mask_arr)
+    mask_instance_count = int(np.count_nonzero(mask_values))
+    if depth_nonzero_fraction <= 0:
+        notes.append("depth has no nonzero values")
+    if mask_instance_count == 0:
+        notes.append("mask has no foreground instances")
+
+    return SampleResult(
+        index=-1,
+        scene_phase=row["scene_phase"],
+        scene_id=row["scene_id"],
+        phase_index=int(row["phase_index"]),
+        difficulty=row["difficulty"],
+        rgb_path=str(rgb_path),
+        depth_path=str(depth_path),
+        mask_path=str(mask_path),
+        rgb_frames=rgb_frames,
+        depth_frames=depth_frames,
+        mask_frames=mask_frames,
+        frame_index=frame_index,
+        rgb_size=rgb.size,
+        depth_size=depth.size,
+        mask_size=mask.size,
+        depth_min=int(depth_arr.min()),
+        depth_max=int(depth_arr.max()),
+        depth_nonzero_fraction=depth_nonzero_fraction,
+        mask_instance_count=mask_instance_count,
+        mask_nonzero_fraction=mask_nonzero_fraction,
+        status="ok" if not notes else "warning",
+        notes=notes,
+    )
+
+
+def make_contact_sheet(
+    repo_root: Path,
+    rows: Iterable[dict[str, str]],
+    out_path: Path,
+    thumb_width: int,
+) -> None:
+    rows = list(rows)
+    cols = 10
+    label_h = 34
+    thumbs: list[tuple[dict[str, str], Image.Image]] = []
+    for row in rows:
+        preview_name = Path(row["image_preview_url"]).name
+        preview_path = repo_root / "preview" / "image_examples" / "preview" / "images" / preview_name
+        image = Image.open(preview_path).convert("RGB")
+        thumb_h = round(image.height * thumb_width / image.width)
+        thumb = image.resize((thumb_width, thumb_h), Image.Resampling.LANCZOS)
+        thumbs.append((row, thumb))
+
+    if not thumbs:
+        raise ValueError("no rows to render")
+    thumb_h = thumbs[0][1].height
+    rows_count = (len(thumbs) + cols - 1) // cols
+    sheet = Image.new("RGB", (cols * thumb_width, rows_count * (thumb_h + label_h)), "white")
+    draw = ImageDraw.Draw(sheet)
+    try:
+        font = ImageFont.truetype("DejaVuSans.ttf", 12)
+    except OSError:
+        font = ImageFont.load_default()
+
+    for i, (row, thumb) in enumerate(thumbs):
+        x = (i % cols) * thumb_width
+        y = (i // cols) * (thumb_h + label_h)
+        sheet.paste(thumb, (x, y))
+        label = f"{i:03d} {row['scene_phase']} {row['difficulty']}"
+        draw.rectangle((x, y + thumb_h, x + thumb_width, y + thumb_h + label_h), fill=(245, 245, 245))
+        draw.text((x + 4, y + thumb_h + 4), label, fill=(0, 0, 0), font=font)
+    out_path.parent.mkdir(parents=True, exist_ok=True)
+    sheet.save(out_path, quality=90)
+
+
+def open_rerun(rrd_path: Path):
+    import rerun as rr
+
+    rr.init("rpx_visual_check", spawn=False)
+    rr.save(str(rrd_path))
+    return rr
+
+
+def log_to_rerun(rr, index: int, row: dict[str, str], rgb: Image.Image, depth: Image.Image, mask: Image.Image) -> None:
+    if hasattr(rr, "set_time_sequence"):
+        rr.set_time_sequence("sample", index)
+    else:
+        rr.set_time("sample", sequence=index)
+    rgb_small = resize_width(rgb.convert("RGB"), 480)
+    depth_arr = np.asarray(resize_width(depth, 480, nearest=True))
+    mask_arr = np.asarray(resize_width(mask, 480, nearest=True))
+    rr.log("rpx/rgb", rr.Image(np.asarray(rgb_small)))
+    try:
+        rr.log("rpx/depth", rr.DepthImage(depth_arr, meter=1000.0))
+    except TypeError:
+        rr.log("rpx/depth", rr.DepthImage(depth_arr))
+    rr.log("rpx/masks", rr.SegmentationImage(mask_arr))
+    markdown = (
+        f"## {index:03d} {row['scene_phase']}\n\n"
+        f"- scene: `{row['scene_id']}`\n"
+        f"- phase: `{row['phase_index']}` / `{row['phase_name']}`\n"
+        f"- difficulty: `{row['difficulty']}`\n"
+        f"- rpx_ds: `{row['rpx_ds']}`\n"
+    )
+    if hasattr(rr, "TextDocument"):
+        rr.log("rpx/metadata", rr.TextDocument(markdown, media_type="text/markdown"))
+
+
+def main() -> None:
+    args = parse_args()
+    repo_root = args.repo_root.resolve()
+    rows_csv = repo_path(repo_root, args.rows_csv)
+    out_dir = repo_path(repo_root, args.out_dir)
+    out_dir.mkdir(parents=True, exist_ok=True)
+    rrd_path = out_dir / "rpx_visual_check.rrd"
+    report_path = out_dir / "visual_check_report.json"
+    contact_sheet_path = out_dir / "rpx_visual_contact_sheet.jpg"
+
+    rows = load_rows(rows_csv, one_per_scene=args.one_per_scene, limit=args.limit)
+    make_contact_sheet(repo_root, rows, contact_sheet_path, args.contact_thumb_width)
+
+    rr = None if args.skip_rerun else open_rerun(rrd_path)
+    results: list[SampleResult] = []
+    for index, row in enumerate(tqdm(rows, desc="checking scene-phases")):
+        rgb_path = repo_path(repo_root, row["source_rgb_shard"])
+        depth_path = repo_path(repo_root, row["source_depth_shard"])
+        mask_path = repo_path(repo_root, row["source_mask_shard"])
+        rgb_names = tar_png_names(rgb_path)
+        frame_index = args.frame_index if args.frame_index is not None else len(rgb_names) // 2
+        rgb, rgb_frames, _ = read_png_from_tar(rgb_path, frame_index)
+        depth, depth_frames, _ = read_png_from_tar(depth_path, frame_index)
+        mask, mask_frames, _ = read_png_from_tar(mask_path, frame_index)
+        result = sample_status(
+            row,
+            rgb,
+            depth,
+            mask,
+            rgb_frames,
+            depth_frames,
+            mask_frames,
+            frame_index,
+            rgb_path,
+            depth_path,
+            mask_path,
+        )
+        result.index = index
+        results.append(result)
+        if rr is not None:
+            log_to_rerun(rr, index, row, rgb, depth, mask)
+
+    warnings = [result for result in results if result.status != "ok"]
+    scene_ids = sorted({result.scene_id for result in results})
+    summary = {
+        "repo_root": str(repo_root),
+        "rows_csv": str(rows_csv),
+        "samples_checked": len(results),
+        "unique_scenes": len(scene_ids),
+        "scene_phase_mode": "one_per_scene" if args.one_per_scene else "all_scene_phases",
+        "rrd_path": str(rrd_path) if rr is not None else None,
+        "contact_sheet_path": str(contact_sheet_path),
+        "warnings": len(warnings),
+        "warning_scene_phases": [result.scene_phase for result in warnings],
+        "results": [asdict(result) for result in results],
+    }
+    report_path.write_text(json.dumps(summary, indent=2), encoding="utf-8")
+    print(f"wrote {report_path}")
+    print(f"wrote {contact_sheet_path}")
+    if rr is not None:
+        print(f"wrote {rrd_path}")
+    if warnings:
+        print(f"warnings: {len(warnings)}")
+    else:
+        print("warnings: 0")
+
+
+if __name__ == "__main__":
+    main()

rerun_check/requirements.txt

@@ -0,0 +1,6 @@
+huggingface_hub
+numpy
+pillow
+pyarrow
+rerun-sdk
+tqdm