3D Hand-Object Pose Estimation from Depth with Convolutional Neural Networks

IEEE International Conference on Automatic Face and Gesture Recognition, 2017, Washington DC

Paper accepted at FG 2017. To the best of our knowledge, we were the first to apply discriminative one-shot methods to the problem of hand-object pose estimation. We use a novel two-stage system to achieve this. We also created a new dataset to achieve this, which we will release publicly soon.

Dataset download: Hand-sphere hand pose estimation, Hand-sphere segmentation, Readme