Mean Shift Rejection: Training Deep Neural Networks Without Minibatch Statistics or Normalization.
Brendan RuffTaylor BeckJoscha BachPublished in: CoRR (2019)
Keyphrases
- mean shift
- neural network
- training process
- object tracking
- training algorithm
- visual tracking
- feature space
- graph cuts
- mode seeking
- target tracking
- neural network training
- sampling procedure
- bilateral filtering
- image segmentation algorithm
- visual object tracking
- density function
- partial occlusion
- particle filter
- kernel density estimation
- kalman filter
- supervised learning
- back propagation
- kernel density
- training set
- object location
- object localisation
- kernel density estimate
- color image segmentation
- region merging
- density estimates
- riemannian manifolds
- appearance model
- segmentation algorithm