LasNack commited on
Commit
e529f8b
·
verified ·
1 Parent(s): e617869

Initial HoloScan release

Browse files
.gitattributes CHANGED
@@ -58,3 +58,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
 
 
 
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
61
+ data/annotations/val.json filter=lfs diff=lfs merge=lfs -text
62
+ data/holocap/holocap_train.json filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ viewer: false
3
+ ---
4
+
5
+ # HoloScan
6
+
7
+ HoloScan is a large-scale 3D scene benchmark introduced for **holo-captioning**, a task that seeks the text equivalent of 3D scenes. It contains more than 15K real and synthetic indoor scenes, including over 13K training scenes, 619 validation scenes, and 83 human-curated test scenes.
8
+
9
+ HoloScan provides structured holo-captions that describe scene entities across four dimensions: semantic tags, spatial locations, entity attributes, and inter-entity relations.
10
+
11
+ For more information, please visit the [HoloCap Project Page](https://visual-ai.github.io/holocap/).
12
+
13
+ ## Data Sources
14
+
15
+ HoloScan is built from four real-world 3D scene datasets and one synthetic dataset:
16
+
17
+ - [ScanNet](https://github.com/scannet/scannet)
18
+ - [3RScan](https://github.com/WaldJohannaU/3RScan)
19
+ - [Matterport3D](https://github.com/niessner/Matterport)
20
+ - [ARKitScenes](https://github.com/apple/ARKitScenes)
21
+ - [Structured3D](https://structured3d-dataset.org/)
22
+
23
+ Before using HoloScan, users must follow the official instructions for each source dataset, including completing any required Terms of Use or accepting the corresponding license, and download the original data from the official providers.
24
+
25
+ Detailed download, directory-layout, and point-cloud preparation instructions are available in the [HoloCap data preparation guide](https://github.com/Visual-AI/HoloCap/blob/main/README_DATA.md).
26
+
27
+ ## Repository Contents
28
+
29
+ This dataset repository provides:
30
+
31
+ - `data/holocap/holocap_train.json`: HoloCap training annotations.
32
+ - `data/annotations/val.json`: validation ground-truth annotations.
33
+ - `data/annotations/test.json`: human-curated test annotations.
34
+ - `data_preparation/metadata/holocap_train.pkl`: training scene metadata and axis-alignment information.
35
+ - `data_preparation/metadata/holocap_val.pkl`: validation scene metadata and axis-alignment information.
36
+ - `data_preparation/metadata/structured3d.pkl`: Structured3D scene metadata and axis-alignment information.
37
+
38
+ This repository does not include raw meshes or aligned PLY files. These files must be downloaded or generated by following the HoloCap data preparation guide.
39
+
40
+ ## Citation
41
+
42
+ If you find HoloScan useful, please cite the HoloCap paper:
43
+
44
+ ```bibtex
45
+ @inproceedings{lin2026holocap,
46
+ title={Holo-Captioning: Toward the Text Equivalent of 3D Scenes},
47
+ author={Lin, Kun-Yu and Bu, Chengke and Li, Zhenguo and Han, Kai},
48
+ booktitle={European Conference on Computer Vision},
49
+ year={2026}
50
+ }
51
+ ```
52
+
53
+ ## License
54
+
55
+ HoloScan annotations and metadata are provided subject to the applicable release terms. The original ScanNet, 3RScan, Matterport3D, ARKitScenes, and Structured3D data remain governed by their respective official licenses and Terms of Use.
data/annotations/test.json ADDED
The diff for this file is too large to render. See raw diff
 
data/annotations/val.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4e6cb778597311180f280a81db2c246be4cf61b077549f8a912ae12388c5a8dc
3
+ size 35710961
data/holocap/holocap_train.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5bb3038b86d55a7c14980b33378b0993d5844e6598ec3b862fe7f5edcccf3391
3
+ size 843323737
data_preparation/metadata/holocap_train.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5fc49cd0b63fa9622b19ca82f81d25cbce43bc5ed4c7da86ddb381e6c30e0c61
3
+ size 590720776
data_preparation/metadata/holocap_val.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d5a719479680aa7823597dd4936ab72ffb725d19e5aef9f50727d4f858f67a5
3
+ size 153185690
data_preparation/metadata/structured3d.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6aa7bd77e92df11cb264e9c1f230d527a86974b9d3ad180bc4e717efa644e1b5
3
+ size 49517956