Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Dataset Card for DuoduoCLIP
|
2 |
+
|
3 |
+
In this data repo we provide the data used in the paper **Duoduo CLIP: Efficient 3D Understanding with Multi-View Images.**
|
4 |
+
The data usage and code can be found in the [github repo](https://github.com/3dlg-hcvc/DuoduoCLIP).
|
5 |
+
|
6 |
+
***Note: We provide the lvis evaluation data in the initial release, we will soon upload the data and process scripts required for training.***
|
7 |
+
|
8 |
+
## Dataset Details
|
9 |
+
|
10 |
+
### Dataset Sources
|
11 |
+
|
12 |
+
Multi-view images of 3D objects were used in training our models.
|
13 |
+
A majority of the Objaverse renderings come from the Zero123 paper (MIT license) listed below.
|
14 |
+
The initial release with LVIS images in 3dlg-hcvc/DuoduoCLIP-data/lvis_split are also mostly from Zero123 with some preprocessing.
|
15 |
+
Note we provide this for ease of evaluation.
|
16 |
+
For the other releases we will provide the dataset preprocessing scripts instead and point users towards their download link.
|
17 |
+
For objects not rendered by Zero123 and the other 3 datasets (ABO, ShapeNet and 3D-FUTURE) we render using Zero123's blender script.
|
18 |
+
We thank the authors for providing their dataset and code!
|
19 |
+
|
20 |
+
1. Zero123
|
21 |
+
|
22 |
+
- **Repository:** https://github.com/cvlab-columbia/zero123
|
23 |
+
- **Paper:** https://arxiv.org/abs/2303.11328
|
24 |
+
|
25 |
+
2. Objaverse
|
26 |
+
|
27 |
+
- **Repository:** https://github.com/allenai/objaverse-xl
|
28 |
+
- **Paper:** https://arxiv.org/abs/2212.08051
|
29 |
+
|
30 |
+
3. ABO
|
31 |
+
|
32 |
+
- **Repository:** https://github.com/jazcollins/amazon-berkeley-objects
|
33 |
+
- **Paper:** https://arxiv.org/abs/2110.06199
|
34 |
+
|
35 |
+
4. ShapeNet
|
36 |
+
|
37 |
+
- **Repository:** https://huggingface.co/ShapeNet
|
38 |
+
- **Paper:** https://arxiv.org/abs/1512.03012
|
39 |
+
|
40 |
+
5. 3D-FUTURE
|
41 |
+
|
42 |
+
- **Website:** https://tianchi.aliyun.com/specials/promotion/alibaba-3d-future
|
43 |
+
- **Paper:** https://arxiv.org/abs/2009.09633
|
44 |
+
|
45 |
+
### Embeddings from our Models
|
46 |
+
|
47 |
+
We will also provide objaverse embeddings produced by our released models in this repo.
|
48 |
+
In the initital release we provide the objaverse embeddings produced by our **Four_1to6F_bs1600_LT6** model.
|
49 |
+
Please see the [model card](https://huggingface.co/3dlg-hcvc/DuoduoCLIP) for more details and [github repo](https://github.com/3dlg-hcvc/DuoduoCLIP) for usage.
|