Datasets:

Modalities:
Image
Languages:
English
Size:
n<1K
ArXiv:
License:
r3ds / README.md
sraychau's picture
Updated readme with annotation detail
5146542 verified
---
license: afl-3.0
language:
- en
pretty_name: 'R3DS: Reality-linked 3D Scenes for Panoramic Scene Understanding'
size_categories:
- n<1K
---
### R3DS: Reality-linked 3D Scenes for Panoramic Scene Understanding
#### ECCV 2024
[Qirui Wu](https://qiruiw.github.io/), [Sonia Raychaudhuri](https://sonia-raychaudhuri.github.io/), [Daniel Ritchie](https://dritchie.github.io/), [Manolis Savva](https://msavva.github.io/), [Angel X. Chang](http://angelxuanchang.github.io/)
[Project](https://3dlg-hcvc.github.io/r3ds/) | [arXiv](https://arxiv.org/abs/2403.12301) | [Data]()
This is the official version of the R3DS dataset. You can request access and proceed to checkout the data:
```
git clone [email protected]:datasets/3dlg-hcvc/rlsd
```
The dataset is structured into different folders:
```
mp3d_arch -- contains a subset of the Matterport3D architecture files that we have used
scenes -- contains the annotated json files for each scene
renderings -- contains various renderings of the annotated scene
original_panoramas -- original panoramas (without annotations) from Matterport3D
annotated_panoramas -- panorama renderings colored by object instance ids
annotated_scene_hierarchy -- hierarchy of the annotated objects for each scene
annotated_topdown -- topdown images of each annotated scene colored by object instance ids
annotated_model_ids.txt -- list of all model ids used to annotate the scenes
```
The annotated scene json has the following structure:
```
{
"format": "sceneState",
"version": "v0.2.0",
"scene": {
"up": ..., # "x", "y" and "z"
"front": ..., # "x", "y" and "z"
"unit": 1,
"assetSource": [
"wayfair",
"3dw"
],
"object": [
{
"id": "a_883",
"modelId": "wayfair.ANDO6573",
"index": 0,
"parentIndex": -1,
"parentId": "1_8_1",
"transform": {
"rows": 4,
"cols": 4,
"data": ... # 1x16 array
},
"obb": {
"centroid": ..., # 1x3 array
"axesLengths": ..., # 1x3 array,
"normalizedAxes": ..., # 1x9 array,
"min": ..., # 1x3 array,
"max": ... # 1x3 array
}
}, ... # multiple object annotations
],
"camera": [
{
"up": ..., # "x", "y" and "z"
"position": ..., # "x", "y" and "z"
"target": ..., # "x", "y" and "z"
"isOrtho": false,
"name": "current"
}
],
"arch": {
"ref": "mp3dArch.1pXnuDYAj8r_L1"
}
},
"selected": [],
"maskObjectAssignments": [
{
"photoId": ..., # same as viewpointId
"maskId": 279,
"clickPoint": [
0.8349366636325659,
0.4788185224119439
],
"objectInstanceId": "a_883"
}, ... # multiple mask2object assignments
],
"maskComments": [ # comments from annotators against missed annotations
{
"reason": "WRONG-CATEGORY",
"comment": "Picture",
"maskId": "304"
}, ... # multiple comments
],
"maskInfos": [
{
"id": 276,
"label": "curtain", # category
"type": "mask",
"pointNormalized": ..., # "x" and "y"
}, ... # multiple mask infos
]
}
```