High-dimensional Motion Segmentation by Variational Autoencoder and Gaussian Processes.
Masatoshi NaganoTomoaki NakamuraTakayuki NagaiDaichi MochihashiIchiro KobayashiWataru TakanoPublished in: IROS (2019)
Keyphrases
- motion segmentation
- gaussian processes
- high dimensional
- optical flow
- model selection
- gaussian process
- hyperparameters
- image sequences
- motion estimation
- moving objects
- motion model
- missing data
- low dimensional
- gaussian process regression
- dimensionality reduction
- spectral clustering
- image segmentation
- multi task
- high dimensional data
- data points
- feature space
- computer vision
- feature points
- cross validation
- probabilistic model
- feature selection
- three dimensional
- similarity measure
- motion field
- data clustering
- multi class