HanningLiu/litereality-reconstruction-comparison
LiteReality Geometry Reconstruction and Scene Understanding Comparisons 本数据集包含 8 个室内场景的两组 Rerun 对比结果：三维几何重建对比，以及 RoomPlan 与 SpatialLM 的场景理解对比。数据集仅发布实验可视化 RRD 与元数据，不包含原始 RGB-D、PLY、视频、RoomPlan USDZ 或模型权重。 This dataset contains two groups of Rerun comparisons for eight indoor scenes: 3D geometry reconstruction and RoomPlan-versus-SpatialLM scene understanding. It publishes experiment recordings and metadata only; raw RGB-D, PLY, video, RoomPlan USDZ, and model weights are not… See the full description on the dataset page: https://huggingface.co/datasets/HanningLiu/litereality-reconstruction-comparison.
LiteReality Geometry Reconstruction and Scene Understanding Comparisons
本数据集包含 8 个室内场景的两组 Rerun 对比结果：三维几何重建对比，以及 RoomPlan 与 SpatialLM 的场景理解对比。数据集仅发布实验可视化 RRD 与元数据，不包含原始 RGB-D、PLY、视频、RoomPlan USDZ 或模型权重。
This dataset contains two groups of Rerun comparisons for eight indoor scenes: 3D geometry reconstruction and RoomPlan-versus-SpatialLM scene understanding. It publishes experiment recordings and metadata only; raw RGB-D, PLY, video, RoomPlan USDZ, and model weights are not included.
Directory layout / 目录结构
scenes/{scene_id}/geometry_comparison.rrd
scene_understanding/{scene_id}/comparison_original_frame.rrdThe existing scenes/ paths are retained for backward compatibility. Scene-understanding recordings use the separate scene_understanding/ tree and do not overwrite geometry recordings.
现有 scenes/ 路径为兼容已有链接而保留；场景理解结果位于独立的 scene_understanding/ 目录，不覆盖几何重建文件。
1. Geometry reconstruction / 几何重建
Each recording compares three point-cloud layers in the original scan frame:
scannerapp_dense: Scanner App dense reconstruction reference / Scanner App 稠密重建参考。open3d_all_depth: Open3D TSDF fused from all accepted depth frames / 全部合格深度帧融合结果。open3d_rgb_matched: Open3D TSDF using RGB-matched frames / RGB 时间匹配帧融合结果。
Pinned geometry data commit: 281c1d32d3d9698bea1ff75afb91e4d38deef099
2. Scene understanding / 场景理解
Each comparison_original_frame.rrd overlays the same three point clouds with four independently switchable layout groups:
roomplan: normalized Apple RoomPlan product baseline / 规范化 RoomPlan 产品基线。scanner_dense_greedy: SpatialLM prediction from Scanner App dense point cloud.open3d_all_depth_greedy: SpatialLM prediction from all-depth Open3D TSDF point cloud.open3d_rgb_matched_greedy: SpatialLM prediction from RGB-matched Open3D TSDF point cloud.
SpatialLM inputs were independently Manhattan-aligned, processed with a 0.025 m preprocessing voxel and 0.025 m model grid, and decoded greedily. Predictions are transformed back into the original scan coordinate frame in these comparison recordings. RoomPlan is a common product baseline, not absolute ground truth.
SpatialLM 的三种输入分别独立完成 Manhattan 对齐，预处理体素和模型网格均为 0.025 m，并使用 greedy 解码。该组 RRD 已将预测转换回原始扫描坐标系。RoomPlan 是统一产品基线，并非绝对真值。
Pinned scene-understanding data commit: ee239cfb9f33bfb870887f71165d53d7fbf88ba4
Rerun version / 版本
All recordings were written with Rerun 0.21.0. Use the matching web viewer version for consistent playback. embed_links.md retains the geometry links; scene_understanding_embed_links.md contains the new scene-understanding links for Notion Embed blocks.
See manifest.json for byte sizes, SHA-256 checksums, recording groups, parameters, and immutable data commits. No license is asserted by this dataset card.