Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,52 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# The test split of Drag-a-Move dataset
|
2 |
+
|
3 |
+
This test set is used to evaluate [DragAPart](https://dragapart.github.io/) and [Puppet-Master](https://vgg-puppetmaster.github.io/).
|
4 |
+
|
5 |
+
### How to use it?
|
6 |
+
|
7 |
+
1. **Frame folders & Ground‑truth videos**
|
8 |
+
Under the dataset root, each sample has its own sub‑folder.
|
9 |
+
Inside each sub‑folder you will find **14 PNG frames** extracted from the ground‑truth video:
|
10 |
+
|
11 |
+
{sample_id}/
|
12 |
+
cXXX_tYYY.png
|
13 |
+
cXXX_tYYY.png
|
14 |
+
…
|
15 |
+
|
16 |
+
where `cXXX` is the zero‑padded camera index and `tYYY` is the zero‑padded frame index.
|
17 |
+
|
18 |
+
2. **Metadata**
|
19 |
+
A `metadata.json` file in the dataset root maps each sample ID to two fields:
|
20 |
+
|
21 |
+
* `reverse` — `true` if the original motion runs backward (larger frame index first), otherwise `false`.
|
22 |
+
* `per_frame_drags` — the sparse motion trajectories as drags. It's an array of shape (14, 5, 4), where it specifies the starting coordinates and ending coordinates of the 5 drags on the 14 frames. The coordinates are normalized to [0, 1] range.
|
23 |
+
|
24 |
+
### Dataset structure
|
25 |
+
|
26 |
+
Drag-a-Move-test-split/
|
27 |
+
├── 00000/
|
28 |
+
│ ├── c031_t000.png
|
29 |
+
│ ├── c031_t010.png
|
30 |
+
│ └── … (14 frames total)
|
31 |
+
├── 00001/
|
32 |
+
│ ├── c005_t020.png
|
33 |
+
│ └── …
|
34 |
+
├── …
|
35 |
+
├── metadata.json
|
36 |
+
└── README.md
|
37 |
+
|
38 |
+
### Citation
|
39 |
+
|
40 |
+
@article{li2024dragapart,
|
41 |
+
title = {DragAPart: Learning a Part-Level Motion Prior for Articulated Objects},
|
42 |
+
author = {Li, Ruining and Zheng, Chuanxia and Rupprecht, Christian and Vedaldi, Andrea},
|
43 |
+
journal = {arXiv preprint arXiv:2403.15382},
|
44 |
+
year = {2024}
|
45 |
+
}
|
46 |
+
|
47 |
+
@article{li2024puppetmaster,
|
48 |
+
title = {Puppet-Master: Scaling Interactive Video Generation as a Motion Prior for Part-Level Dynamics},
|
49 |
+
author = {Li, Ruining and Zheng, Chuanxia and Rupprecht, Christian and Vedaldi, Andrea},
|
50 |
+
journal = {arXiv preprint arXiv:2408.04631},
|
51 |
+
year = {2024}
|
52 |
+
}
|