Hand Graph Cnn

3D Hand Shape and Pose Estimation from a Single RGB Image
Alternatives To Hand Graph Cnn
Project NameStarsDownloadsRepos Using ThisPackages Using ThisMost Recent CommitTotal ReleasesLatest ReleaseOpen IssuesLicenseLanguage
Hand Graph Cnn388
3 years ago13otherPython
3D Hand Shape and Pose Estimation from a Single RGB Image
3 years agomit
3D Hand Shape and Pose Estimation from a Single RGB Image
2 years agomitPython
Code for "Learning Unsupervised Hierarchical Part Decomposition of 3D Objects from a Single RGB Image", CVPR 2020
4 years ago2otherPython
Repository for Text2Shape
4 years ago2MATLAB
3 years ago3mpl-2.0C++
Use MXNet to accelerated Image-Processing in VapourSynth.
Owesome 3d Semantic Segmentation Reconstruction14
5 years ago
4 years agoPython
3D Shape Generation Baselines in PyTorch.
3 years ago15September 30, 20201JavaScript
An open source JavaScript library for processing image using open cv
3d Machine Learning Complete Resources7
5 years ago
Alternatives To Hand Graph Cnn
Select To Compare

Alternative Project Comparisons

3D Hand Shape and Pose Estimation from a Single RGB Image

Open source of our CVPR 2019 paper "3D Hand Shape and Pose Estimation from a Single RGB Image"

prediction example


This work is based on our CVPR 2019 paper. You can also check our project webpage and supplementary video for a deeper introduction.

This work addresses a novel and challenging problem of estimating the full 3D hand shape and pose from a single RGB image. Most current methods in 3D hand analysis from monocular RGB images only focus on estimating the 3D locations of hand keypoints, which cannot fully express the 3D shape of hand. In contrast, we propose a Graph Convolutional Neural Network (Graph CNN) based method to reconstruct a full 3D mesh of hand surface that contains richer information of both 3D hand shape and pose. To train networks with full supervision, we create a large-scale synthetic dataset containing both ground truth 3D meshes and 3D poses. When fine-tuning the networks on real-world datasets without 3D ground truth, we propose a weakly-supervised approach by leveraging the depth map as a weak supervision in training. Through extensive evaluations on our proposed new datasets and two public datasets, we show that our proposed method can produce accurate and reasonable 3D hand mesh, and can achieve superior 3D hand pose estimation accuracy when compared with state-of-the-art methods.


If you find our work useful in your research, please consider citing:

  title={3D Hand Shape and Pose Estimation from a Single RGB Image},
  author={Ge, Liuhao and Ren, Zhou and Li, Yuncheng and Xue, Zehao and Wang, Yingying and Cai, Jianfei and Yuan, Junsong},


  1. Install pytorch >= v0.4.0 following official instruction.
  2. Clone this repo, and we'll call the directory that you cloned as ${HAND_ROOT}.
  3. Install dependencies:
    pip install -r requirements.txt

Running the code

  1. Evaluate on our real-world dataset and visualize the results of hand mesh and pose.

    python eval_script.py --config-file "configs/eval_real_world_testset.yaml"

    The visualization results will be saved to ${HAND_ROOT}/output/configs/eval_real_world_testset.yaml/

  2. Evaluate on STB dataset.

    Download STB dataset to ${HAND_ROOT}/data/STB.

    Run the following script:

    python eval_script.py --config-file "configs/eval_STB_dataset.yaml"

    The pose estimation results will be saved to ${HAND_ROOT}/output/configs/eval_STB_dataset.yaml/pose_estimations.mat

3D hand shape and pose dataset

We release the 3D hand shape and pose dataset. It contains a large scale synthetic image dataset for training and validation, and a small real-world image dataset for testing. For details, please go to the data folder in this repository.

Popular Rgb Projects
Popular Shape Projects
Popular User Interface Categories

Get A Weekly Email With Trending Projects For These Categories
No Spam. Unsubscribe easily at any time.
3d Graphics
Human Pose Estimation