Improve metadata and add technical documentation
Browse filesHi! I'm Niels from the Hugging Face community science team.
I've opened this PR to improve the model card's discoverability and technical documentation:
- Added `objaverse` to the `datasets` metadata field.
- Added descriptive tags (`relighting`, `multi-view`, `reconstruction`) to help users find the model.
- Included "Input Format" and "Outputs" sections from the official GitHub repository to provide clearer guidance on using the model.
- Updated the Quickstart command with parameters found in the source repository for better parity.
These changes ensure the repository is well-documented for the community!
README.md
CHANGED
|
@@ -3,8 +3,14 @@ license: other
|
|
| 3 |
license_name: stabilityai-ai-community
|
| 4 |
license_link: LICENSE.md
|
| 5 |
pipeline_tag: image-to-3d
|
|
|
|
|
|
|
| 6 |
tags:
|
| 7 |
- image-to-3d
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
- checkpoint
|
| 9 |
inference: false
|
| 10 |
---
|
|
@@ -34,6 +40,27 @@ Please note: For individuals or organizations generating annual revenue of US $1
|
|
| 34 |
* **Project page**: https://reli3d.jdihlmann.com/
|
| 35 |
* **arXiv paper**: https://arxiv.org/abs/2603.19753
|
| 36 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
### Files
|
| 38 |
|
| 39 |
* **config.yaml**: ReLi3D inference config
|
|
@@ -64,6 +91,8 @@ python demos/reli3d/infer_from_transforms.py \
|
|
| 64 |
--input-root demo_files/objects \
|
| 65 |
--objects Camera_01 \
|
| 66 |
--output-root outputs \
|
|
|
|
|
|
|
| 67 |
--overwrite
|
| 68 |
```
|
| 69 |
|
|
@@ -107,4 +136,4 @@ Please report any issues with the model or contact us:
|
|
| 107 |
* Security issues: security@stability.ai
|
| 108 |
* Privacy issues: privacy@stability.ai
|
| 109 |
* License and general: https://stability.ai/license
|
| 110 |
-
* Enterprise license: https://stability.ai/enterprise
|
|
|
|
| 3 |
license_name: stabilityai-ai-community
|
| 4 |
license_link: LICENSE.md
|
| 5 |
pipeline_tag: image-to-3d
|
| 6 |
+
datasets:
|
| 7 |
+
- objaverse
|
| 8 |
tags:
|
| 9 |
- image-to-3d
|
| 10 |
+
- 3d
|
| 11 |
+
- reconstruction
|
| 12 |
+
- relighting
|
| 13 |
+
- multi-view
|
| 14 |
- checkpoint
|
| 15 |
inference: false
|
| 16 |
---
|
|
|
|
| 40 |
* **Project page**: https://reli3d.jdihlmann.com/
|
| 41 |
* **arXiv paper**: https://arxiv.org/abs/2603.19753
|
| 42 |
|
| 43 |
+
### Technical Details
|
| 44 |
+
|
| 45 |
+
#### Input Format
|
| 46 |
+
Each object directory must contain a camera specification and RGBA frames:
|
| 47 |
+
```text
|
| 48 |
+
input_root/
|
| 49 |
+
object_name/
|
| 50 |
+
transforms.json
|
| 51 |
+
rgba/
|
| 52 |
+
0000.png
|
| 53 |
+
0010.png
|
| 54 |
+
...
|
| 55 |
+
```
|
| 56 |
+
Expected `transforms.json` frame keys include `file_path`, `transform_matrix` (or `camera_transform`), and `camera_fov`.
|
| 57 |
+
|
| 58 |
+
#### Outputs
|
| 59 |
+
The model produces the following per object:
|
| 60 |
+
- `mesh.glb`: UV-unwrapped textured mesh.
|
| 61 |
+
- `illumination.hdr`: Estimated environment illumination (if predicted).
|
| 62 |
+
- `run_info.json`: Metadata about the inference run.
|
| 63 |
+
|
| 64 |
### Files
|
| 65 |
|
| 66 |
* **config.yaml**: ReLi3D inference config
|
|
|
|
| 91 |
--input-root demo_files/objects \
|
| 92 |
--objects Camera_01 \
|
| 93 |
--output-root outputs \
|
| 94 |
+
--num-views 4 \
|
| 95 |
+
--texture-size 256 \
|
| 96 |
--overwrite
|
| 97 |
```
|
| 98 |
|
|
|
|
| 136 |
* Security issues: security@stability.ai
|
| 137 |
* Privacy issues: privacy@stability.ai
|
| 138 |
* License and general: https://stability.ai/license
|
| 139 |
+
* Enterprise license: https://stability.ai/enterprise
|