Need help with 3dhand?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

boukhayma
147 Stars 23 Forks 11 Commits 10 Opened issues

Services available

!
?

Need anything else?

Contributors list

No Data

3D Hand Shape and Pose from Images in the Wild

Adnane Boukhayma, Rodrigo de Bem, Philip H.S. Torr. CVPR 2019 (Oral)

PCK curves

We provide scripts and data to plot 3D & 2D PCK curves in figures 4,5,6,7,8 of the paper in directory

PCK
. We use gnuplot for plotting the figures. For example:
cd PCK/dataset
./figx
where
dataset
is one of
{stereo, egodexter, dexter+object, mpii+nzsl}
and
figx
is one of
{fig4,5,6,7,8}
.

Requirements

  • Python 2.7
  • PyTorch 0.3
  • MANO

Data preprocessing

Input images are assumed to be crops of fixed size around the hand. Using this command, we find the tightest rectangular box of edge size l containing the hand. Images are then cropped with a square patch of size 2.2xl. Input images are located in

data/original
and cropped images are saved in
data/cropped
.
python crop.py
This command generates the 2D joint heat map images used as additional input to our network. You will need to build PyOpenPose.
python heatmap.py
This command generates hand image segmentations using 2D joint locations and GrabCut.
python segment.py

Testing

Download the model weights. Put model-0.pth and model-1.pth in directory

data
.
https://drive.google.com/file/d/1Ar9gUuq1kpVUUOlGsuOqfeiyv2GdixS9
https://drive.google.com/file/d/135Ka6XQVcFBGTLyVu1D0dzXOcc-0XNao
Run the following command to test our method on cropped image examples in
data/cropped
.
python tester.py
Our resnet+Mano model is implemented in
model.py
.

Synthetic data creation

This command colors registered meshes in

data/meshes_registered
using their colored corresponding unregistered meshes in
data/meshes_unregistered
. Results are saved in
data/meshes_colored
.
python create_colored_meshes.py
Using mesh colour instances in
data/meshes_colored
, background images in
data/backgrounds
, and sampling global poses and geometries from MANO, this command creates synthetic image examples.
python create_synthetic_data.py

Citation

@inproceedings{boukhayma20193d,
  title={3d hand shape and pose from images in the wild},
  author={Boukhayma, Adnane and Bem, Rodrigo de and Torr, Philip HS},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  pages={10843--10852},
  year={2019}
}

Acknowledgement

This work was supported by the ERC grant ERC-2012-AdG 321162-HELIOS, the EPSRC grant Seebibyte EP/M013774/1 and the EPSRC/MURI grant EP/N019474/1.

License

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.