Deep Optical Flow Supervised Learning With Prior Assumptions.
Xuezhi XiangMingliang ZhaiRongfang ZhangYulong QiaoAbdulmotaleb El-SaddikPublished in: IEEE Access (2018)
Keyphrases
- image motion
- optical flow
- supervised learning
- flow field
- image sequences
- camera motion
- motion field
- motion model
- motion parameters
- unsupervised learning
- motion estimation
- motion analysis
- least squares
- optical flow computation
- prior knowledge
- training data
- computer vision
- optical flow estimation
- statistical learning
- semi supervised
- active learning
- learning algorithm
- vector field
- learning problems
- optical flow fields
- reinforcement learning
- training samples
- class labels
- labeled data
- feature tracking
- moving objects
- qualitative and quantitative evaluation
- dense optical flow
- neural network
- estimation of optical flow
- smoothness constraint
- multiscale
- machine learning
- motion segmentation
- learning tasks
- deep learning
- structure tensor
- d scene
- training examples
- training set
- rigid body motion