报告题目：3D Hand Pose Regression
3D hand pose regression is a fundamental component in many modern HCI applications such as sign language recognition, virtual object manipulation, game control, etc. The problem has many challenges including high degrees of freedom, severe viewpoint changes, self-occlusion and sensor noise. This talk covers two different discriminative methods that are designed for 3D hand pose regression, as well as some insights that can be generalised to other computer vision problems.
Traditional discriminative methods often fall into two categories: holistic and patch-based. Holistic methods are efficient but less flexible due to their nearest-neighbour nature. Patch-based methods can better generalise to unseen samples by considering local appearance only. However, they are complex because each pixel need to be classified or regressed during testing. In contrast to these two baselines, we proposes Latent Regression Forest (LRF), a method that models the pose estimation problem as a coarse-to-fine search. This inherently combines the efficiency of a holistic method and the flexibility of a patch-based method, and thus results in over 60FPS without CPU/GPU optimisation. Nonetheless, the hierarchical design of LRF makes it difficult to verify the intermediate results, hence errors are accumulating. Targeting these drawbacks, Hierarchical Sampling Forest (HSF) is proposed to model this problem as a progressive search, guided by kinematic structure. Hence the intermediate results (partial poses) can be verified by a new efficient energy function. Consequently more accurate full pose results can be produced.
Danhang Tang, Hyung Jin Chang, Alykhan Tejani, Tae-Kyun Kim
Latent Regression Forest: Structured Estimation of 3D Articulated Hand Posture,
Trans on Pattern Analysis and Machine Intelligence (TPAMI), accepted to appear, 2016.
Danhang Tang, Jonathan Taylor, Pushmeet Kohli, Cem 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).
Danhang Tang, Hyung Jin Chang*, Alykhan 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).*indicates equal contribution.
Danhang Tang, Tsz Ho 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).
唐丹航，于2015年底获得英国帝国理工的博士学位。博士研究课题围绕三维手势识别，从机器学习的角度进行了多种角度的探索。读博期间的成果发表在了PAMI等顶级期刊，以及CVPR、ICCV和ECCV等机器视觉领域的顶级会议（多篇为Oral Presentation）。唐丹航曾经于2007至2009年为由NASA科学家创立的公司Evryx Ltd. 担任中国分公司技术负责人，期间参与了世界上第一个手机图像识别软件SnapNow的研发工作，并参与起草了中国电信的可视化搜索规范。读博期间曾经在微软剑桥研究院在Kinect人体识别算法的发明人Jamie Shotton指导下实习，并于毕业后继续在微软雷德蒙德研究院访问。唐丹航现为旧金山创业公司perceptiveIO Inc.的高级科学家。该公司大部分成员来自微软研究院，曾经为Kinect以及Hololens提供了核心视觉算法。
GMT+8, 2017-4-28 10:28 , Processed in 0.030506 second(s), 17 queries .