图 书 馆
动态二进制潜变量模型的三维人体姿态跟踪Dynamical Binary Latent Variable Models for 3D Human Pose Tracking |
|
| 课程网址: | http://videolectures.net/cvpr2010_taylor_dblv/ |
| 主讲教师: | Graham Taylor |
| 开课单位: | 圭尔夫大学 |
| 开课时间: | 2010-07-19 |
| 课程语种: | 英语 |
| 中文简介: |  |
| 课程简介: | We introduce a new probability like latent variable model, which is called the conditional constrained Boltzmann implicit hybrid (imcrbm) for human posture tracking. The main performances of imcrbm are as follows: (1) learning is the learning that the number of linear training samples can be learned from large datasets; (2) learning multiple activity coherent models; (3) it will automatically discover the "moves" of atoms ”; and (4) can infer the transition between activities even when the transition does not appear in the training set. We describe the model and how it is known that we prove its use in context of Bayesian filtering for multi view and single eye attitude tracking. The model deals with difficult scenarios, including multiple activities and transitions between them. We report on the state-of-the-art results of the humaneva dataset. |
| 关 键 词: | 类概率; 潜变量模型; 贝叶斯滤波 |
| 课程来源: | 视频讲座网 |
| 最后编审: | 2021-09-15：zyk |
| 阅读次数: | 63 |