UniSerj-CSDN博客

原创 HP-GAN: Probabilistic 3D human motion prediction via GAN 论文解读

论文: HP-GAN: Probabilistic 3D human motion prediction via GAN论文地址:https://arxiv.org/pdf/1711.09561.pdf作者的模型是大多人体运动预测都会用到的sequence-to-sequence model，但是在训练过程中给定相同的输入，通过GAN网络学习多种可能的未来姿势，即通过深度神经网络做概率性运...

2018-08-03 08:54:50 1263 1

原创 IJCAI 2018 Long-Term Human Motion Prediction by Modeling Motion Context and Enhancing Motion Dynamic

论文: Long-Term Human Motion Prediction by Modeling Motion Context and Enhancing Motion Dynamic论文地址: https://arxiv.org/pdf/1805.02513.pdf作者针对很多模型长期预测方面的不足，提出了一种新的模型改进人体运动的长期预测。通过motion context总结之前所有...

2018-08-01 18:32:39 1402

原创 CVPR 2016 Structural-RNN: Deep Learning on Spatio-Temporal Graphs 论文解读

论文: Structural-RNN: Deep Learning on Spatio-Temporal Graphs论文地址: https://arxiv.org/pdf/1511.05298.pdf作者提出一种基于st-graph的RNN混合模型Structural-RNN（S-RNN），该模型特点是捕捉人体与周围环境中物体之间的相关性从而对人物未来的运动对预测。不用于其他模型针对mo...

2018-07-31 18:47:09 2711

原创 ICCV 2015 Recurrent Network Models for Human Dynamics 论文解读

论文: Recurrent Network Models for Human Dynamics论文地址: https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Fragkiadaki_Recurrent_Network_Models_ICCV_2015_paper.pdf通过对mocap数据集以及视频的学习，...

2018-07-30 19:05:28 1304

原创 CVPR 2018 Convolutional Sequence to Sequence Model for Human Dynamics 论文解读

论文: Convolutional Sequence to Sequence Model for Human Dynamics论文地址: https://arxiv.org/abs/1805.00655通过对mocap数据的学习，基于作者提出的新模型Convolutional Sequence to Sequence Model，实现对于人体运动的后几帧预测，对比之前基于RNN的resid...

2018-07-27 12:59:14 1339 2

原创 CVPR 2017 On human motion prediction using recurrent neural networks 论文解读

论文: On human motion prediction using recurrent neural networks论文地址: https://arxiv.org/abs/1705.02445通过对mocap数据的学习，基于recurrent neural network提出新的模型，实现对于人物运动的后几帧预测，相比于LSTM-3LR（3 layers of LSTM），ERD...

2018-07-25 18:37:41 2844 1

原创 CVPR 2017 Deep representation learning for human motion prediction and classification 论文解读

论文: Deep representation learning for human motion prediction and classification论文地址: https://arxiv.org/abs/1702.07486通过对mocap数据的学习，基于三种不同的 temporal encoder structures，实现对于人物运动的后几帧预测Table of Con...

2018-07-24 10:55:13 1437

UniSerj的博客