coolho1129 commited on
Commit
b10fdd9
·
verified ·
1 Parent(s): 0773a48

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +65 -0
README.md CHANGED
@@ -1,3 +1,68 @@
1
  ---
2
  license: cc-by-nc-4.0
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: cc-by-nc-4.0
3
+ pipeline_tag: image-to-3d
4
+ tags:
5
+ - hand-avatar
6
+ - gaussian-splatting
7
+ - 3d-reconstruction
8
+ - mano
9
+ - neurips-2026
10
  ---
11
+
12
+ # InstantHand
13
+
14
+ Pretrained weights for **InstantHand: Feed-Forward 3D Gaussian Splatting for Animatable Hand Avatar Reconstruction from a Single Image** (NeurIPS 2026).
15
+
16
+ InstantHand constructs an animatable 3D Gaussian hand avatar from a single image, given off-the-shelf MANO, camera, and mask estimates, without test-time optimization.
17
+
18
+ - **Code:** https://github.com/coolho1129/InstantHand
19
+ - **Project page:** https://coolho1129.github.io/InstantHand-page
20
+
21
+ ## Files
22
+
23
+ Both checkpoints give identical results. Use either one.
24
+
25
+ | File | Size | DINOv3 backbone | Notes |
26
+ |---|---|---|---|
27
+ | `checkpoints/instanthand_full.safetensors` | 1.4 GB | included | No separate DINOv3 download needed |
28
+ | `checkpoints/instanthand_light.safetensors` | 0.2 GB | not included | Requires the DINOv3 ViT-L/16 weights (see the GitHub README) |
29
+ | `DINOv3_LICENSE.md` | | | License of the DINOv3 backbone included in the full checkpoint |
30
+
31
+ ## Usage
32
+
33
+ Download one checkpoint and save it where the code expects it:
34
+
35
+ ```bash
36
+ pip install -U huggingface_hub
37
+ hf download coolho1129/InstantHand checkpoints/instanthand_full.safetensors --local-dir .
38
+ mkdir -p exps/neurips2026/stage2_best
39
+ mv checkpoints/instanthand_full.safetensors exps/neurips2026/stage2_best/model.safetensors
40
+ ```
41
+
42
+ Then follow the evaluation instructions in the [GitHub repository](https://github.com/coolho1129/InstantHand), for example:
43
+
44
+ ```bash
45
+ python eval_interhand.py
46
+ python eval_hanco.py
47
+ ```
48
+
49
+ The **MANO** hand model is not redistributed here. Register at https://mano.is.tue.mpg.de and follow the GitHub README.
50
+
51
+ ## Training data
52
+
53
+ The model was trained on InterHand2.6M, HO3Dv3, Goliath, Hand Appearance (HARP), and PALM, together with synthetic appearance data generated following Zhao et al. (CVPR 2025). The InterHand2.6M evaluation subject and all HanCo subjects are unseen during training. See Appendix B of the paper for details.
54
+
55
+ ## License
56
+
57
+ The InstantHand weights are released under **CC BY-NC 4.0** for non-commercial research use, consistent with the licenses of the training datasets. The DINOv3 backbone included in the full checkpoint is distributed under the DINOv3 License (`DINOv3_LICENSE.md`). Use of MANO is subject to the MANO license.
58
+
59
+ ## Citation
60
+
61
+ ```bibtex
62
+ @inproceedings{kim2026instanthand,
63
+ title = {{InstantHand}: Feed-Forward {3D} {Gaussian} Splatting for Animatable Hand Avatar Reconstruction from a Single Image},
64
+ author = {Kim, Chanho and Ali, Muhammad Salman and Gwak, Kyeonghwan and Shaheryar, Muhammad and Park, Incheol and On, Jeongwan and Baek, Seungryul},
65
+ booktitle = {NeurIPS},
66
+ year = {2026}
67
+ }
68
+ ```