File size: 1,454 Bytes
0149420
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
# Dataset Sources and Attribution

Worldscape-MoE is a curated multi-control training collection. Files are
renamed with stable hashes and metadata paths are rewritten relative to the
dataset root.

## Camera control

The released camera subset contains 5,000 RealEstate10K (RE10K) samples paired
with processed 81-frame camera trajectories.

- Project: https://google.github.io/realestate10k/
- Paper: https://arxiv.org/abs/1810.00444

Please cite RealEstate10K when using the camera subset and follow the upstream
dataset's source and usage terms.

## Dual-arm control

The dual-arm subset contains robot videos and aligned 14-dimensional joint
actions curated for Worldscape-MoE.

## Hand/action-map control

The action-map subset contains target videos paired with dense interaction
control videos curated for Worldscape-MoE.

## LIBERO single-arm control

The LIBERO subset contains deterministic training windows derived from LIBERO
robot-manipulation trajectories.

- Project: https://github.com/Lifelong-Robot-Learning/LIBERO
- Paper: https://arxiv.org/abs/2306.03310

LIBERO project attribution is retained for this derived subset.

## Worldscape-MoE release

The released selection, metadata normalization, path rewriting, deduplication,
and action-window construction were produced for Worldscape-MoE by
EmbodiedCity. Cite the Worldscape-MoE repository when using this packaged
collection:

https://github.com/EmbodiedCity/Worldscape-MoE.code