Datasets:
lemonaddie
commited on
Commit
•
f9acd95
1
Parent(s):
f286927
Upload 3 files
Browse files- README.md +70 -0
- imgs/dataset.png +3 -0
- imgs/vis_objstraj.png +3 -0
README.md
ADDED
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- en
|
4 |
+
---
|
5 |
+
# 360°-Motion Dataset
|
6 |
+
|
7 |
+
[Project page](http://fuxiao0719.github.io/projects/3dtrajmaster) | [Paper](https://drive.google.com/file/d/111Z5CMJZupkmg-xWpV4Tl4Nb7SRFcoWx/view)
|
8 |
+
|
9 |
+
![image/png](imgs/dataset.png)
|
10 |
+
|
11 |
+
### News
|
12 |
+
- [2024-12] We release the V1 dataset (50 entities and 36,000 videos).
|
13 |
+
|
14 |
+
### Data structure
|
15 |
+
|
16 |
+
```
|
17 |
+
├── 360Motion-Dataset Video Number Cam-Obj Distance (m)
|
18 |
+
├── Desert (`desert`) 18,000
|
19 |
+
├── location_data.json
|
20 |
+
├── HDRI [3.43, 13.01]
|
21 |
+
├── loc1 (`snowy street`) 3,600
|
22 |
+
├── loc2 (`park`) 3,600
|
23 |
+
├── loc3 (`indoor open space`) 3,600
|
24 |
+
├── loc11 (`gymnastics room`) 3,600
|
25 |
+
├── loc13 (`autumn forest`) 3,600
|
26 |
+
├── location_data.json
|
27 |
+
├── RefPic
|
28 |
+
├── CharacterInfo.json
|
29 |
+
├── Hemi12_transforms.json
|
30 |
+
```
|
31 |
+
|
32 |
+
> **(1) Released Dataset Information**
|
33 |
+
|
34 |
+
| Argument | Description |Argument | Description |
|
35 |
+
|-------------------------|-------------|-------------------------|-------------|
|
36 |
+
| **Video Resolution** | 480×720 | **Frames/Duration/FPS** | 99/3.3s/30 |
|
37 |
+
| **UE Scenes** | 6 (1 desert+5 HDRIs) | **Video Samples** | 36,000 |
|
38 |
+
| **Hemi12_transforms.json** | 12 surrounding cameras | **CharacterInfo.json** | entity prompts |
|
39 |
+
| **RefPic** | 50 animals | **1/2/3 Trajectory Templates** | 36/60/35 (121 in total) |
|
40 |
+
| **{D/N}_{locX}** | {Day/Night}_{LocationX} | **{C}_ {XX}_{35mm}** | {Close-Up Shot}_{Cam. Index(1-12)} _{Focal Length}|
|
41 |
+
|
42 |
+
|
43 |
+
> **(2) Difference with the Dataset to Train on Our Internal Video Diffusion Model**
|
44 |
+
|
45 |
+
The release of the full dataset regarding more entities and UE scenes is 1) still under our internal license check, 2) awaiting the paper decision.
|
46 |
+
|
47 |
+
| Argument | Released Dataset | Our Internal Dataset|
|
48 |
+
|-------------------------|-------------|-------------------------|
|
49 |
+
| **Video Resolution** | 480×720 (re-rendered) | 384×672 |
|
50 |
+
| **Entities** | 50 (all animals) | 70 (20 humans+50 animals) |
|
51 |
+
| **Video Samples** | 36,000 | 54,000 |
|
52 |
+
| **Scenes** | 6 | 9 (+city, forest, asian town) |
|
53 |
+
| **Trajectory Templates** | 121 | 96 |
|
54 |
+
|
55 |
+
> **(3) Load Dataset Sample**
|
56 |
+
|
57 |
+
1. Change root path to `dataset`. We provide a script to load our dataset (video & entity & pose sequence) as follows. It will generate the sampled video for visualization in the same folder path.
|
58 |
+
|
59 |
+
```bash
|
60 |
+
python load_dataset.py
|
61 |
+
```
|
62 |
+
|
63 |
+
2. Visualize the 6DoF pose sequence via Open3D as follows.
|
64 |
+
|
65 |
+
```bash
|
66 |
+
python vis_trajecotry.py
|
67 |
+
```
|
68 |
+
After running the visualization script, you will get an interactive window like this.
|
69 |
+
|
70 |
+
<img src="imgs/vis_objstraj.png" width="350" />
|
imgs/dataset.png
ADDED
Git LFS Details
|
imgs/vis_objstraj.png
ADDED
Git LFS Details
|