3D Articulated Hand Pose Estimation with Single Depth Images

S. Yuan, Q. Ye, B. Stenger, S. Jain, T-K. Kim
Big Hand 2.2M Benchmark: Hand Pose Data Set and State of the Art Analysis,Proc. of IEEE Conf. on Computer Vision and Pattern Recognition, (CVPR), Honolulu, Hawaii, USA, 2017.

Download: PDF
Q. Ye*, S. Yuan*, T-K. Kim
Spatial Attention Deep Net with Partial PSO for Hierarchical Hybrid Hand Pose Estimation,Proc. of European Conf. on Computer Vision (ECCV), Amsterdam, Netherlands, 2016. (*indicates equal contribution).

Download: PDF
Hyung Jin Chang, Guillermo Garcia-Hernando, Danhang Tang, Tae-Kyun Kim
Spatio-Temporal Hough Forest for Efficient Detection-Localisation-Recognition of Fingerwriting in Egocentric Camera,
Computer Vision and Image Understanding (CVIU) 2016.

Download: PDF
Danhang Tang, Jonathan Taylor, Pushmeet Kohli, Çem Keskin, T-K. Kim, Jamie Shotton
Opening the Black Box: Hierarchical Sampling Optimization for Estimating Human Hand Pose, Proc. of IEEE Int. Conf. on Computer Vision (ICCV), Santiago, Chile, 2015 (oral, acceptance rate=3.3%).

Download: PDF [Supplementary] [Demo on YouTube]
Y. Jang, S. Noh, H. Chang, T-K. Kim, W. Woo
3D Finger CAPE: Clicking Action and Position Estimation under Self-Occlusions in Egocentric Viewpoint, Proc. of IEEE Virtual Reality (VR), Arles, France, 2015 (full paper, accept rate=13.8% (13/94)), also in IEEE Trans. on Visualization and Computer Graphics, 21(4):501-510, Apr 2015.
Download: pdf [Project page]
D. Tang, H.J. Chang*, A. Tejani*, T-K. Kim
Latent Regression Forest: Structured Estimation of 3D Hand Posture, Proc. of IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Columbus, Ohio, USA, 2014 (oral, accept rate=5.75%).
*indicates equal contribution.

Download: PDF [ 960 KB] [Demo on YouTube video] [Project page]
D. Tang, T.H. Yu and T-K. Kim
Real-time Articulated Hand Pose Estimation using Semi-supervised Transductive Regression Forests, Proc. of IEEE Int. Conf. on Computer Vision (ICCV), Sydney, Australia, 2013 (oral, accept rate=2.7%).

Download: PDF [ 1,465 KB] [Demo on YouTube video] [Slides]
Download: New Annotation Training / Testing 


Download: Training / Testing / Our results


Label description:
  • Each line is corresponding to one image.
  • Each line has 16x3 numbers, which indicates (x, y, z) of 16 joint locations. Note that these are joint CENTRE locations.
  • Note that (x, y) are in pixels and z is in mm.
  • The order of 16 joints is Palm, Thumb root, Thumb mid, Thumb tip, Index root, Index mid, Index tip, Middle root, Middle mid, Middle tip, Ring root, Ring mid, Ring tip, Pinky root, Pinky mid, Pinky tip.
  • We used Intel Creative depth sensor. Calibration parameters can be obtained as in Page 119 of SDK Manual:

Many thanks to Guillermo Garcia for help with publishing this dataset.
Many thanks to Danhang Tang, Shanxin Yuan, and Qi Ye for help with reannotating this dataset.