Abstract:Human motion prediction is an increasingly interesting topic in computer vision and robotics. In this paper, we propose a new end-to-end feedforward network, TrajectoryCNN, to predict future poses. Compared with the most existing methods, we introduce a new trajectory space and focus on modeling motion dynamics of the input sequence with coupled spatio-temporal features, dynamic local-global features, and global temporal co-occurrence features in the new space. Specifically, the coupled spatio-temporal features describe the spatial and temporal structural information hidden in a natural human motion sequence, which can be easily mined using CNN by simultaneously covering the spatial and temporal dimensions of the sequence with the convolutional filters. The dynamic local-global features encode different correlations among joint trajectories of human motion (i.e. strong correlations among joint trajectories of one part and weak correlations among joint trajectories of different parts), which can be captured by stacking multiple residual trajectory blocks and incorporating our skeletal representation. The global temporal co-occurrence features represent different importance of different input poses to mine the motion dynamics for predicting future poses, which can be obtained automatically by learning free parameters for each pose with our TrajectoryCNN. Finally, we predict future poses with the captured motion dynamic features in a non-recursive manner. Extensive experiments show that our method achieves state-of-the-art performance on five benchmarks (e.g. Human3.6M, CMU-Mocap, 3DPW, G3D, and FNTU), which demonstrates the effectiveness of our proposed method. The code is available at https://github.com/lily2lab/TrajectoryCNN.git.

Detection of Dummy Trajectories Using Convolutional Neural Networks.

Dummy trajectory generation scheme based on generative adversarial networks

Synthetic Trajectory Generation Through Convolutional Neural Networks

Gps-Simulated Trajectory Detection

Trajectory Forecasting with Neural Networks: an Empirical Evaluation and A New Hybrid Model

A Novel Trajectory Prediction Method Based on CNN, BiLSTM, and Multi-Head Attention Mechanism

MT-CNN: A Lightweight Spatial-Temporal Convolutional Neural Network for Deep Learning of Complex Trajectory Distributions based on Area Partitioning

Detecting Suspects by Large-Scale Trajectory Patterns in the City

Probability Trajectory: One New Movement Description for Trajectory Prediction

Trajectory Recognition of Unmanned Aerial Vehicles Based on LSTM Network

Classification and regression of trajectories rendered as images via 2D Convolutional Neural Networks

MSTCNN: multi-modal spatio-temporal convolutional neural network for pedestrian trajectory prediction

Fuzzy Spatiotemporal Representation Model for Human Trajectory Classification

TrajectoryCNN: A New Spatio-Temporal Feature Learning Network for Human Motion Prediction

LSTM-TrajGAN: A Deep Learning Approach to Trajectory Privacy Protection

Pedestrian trajectory prediction with convolutional neural networks

Motion Trajectory Prediction Based on a CNN-LSTM Sequential Model

DSTCNN: Deformable Spatial-Temporal Convolutional Neural Network for Pedestrian Trajectory Prediction

Context-Aware Trajectory Anomaly Detection

TrajectoryNet: An Embedded GPS Trajectory Representation for Point-based Classification Using Recurrent Neural Networks