Squeeze-SegNet: a new fast deep convolutional neural network for semantic segmentation.
Géraldin NanfackAzeddine ElhassounyRachid Oulad Haj ThamiPublished in: ICMV (2017)
Keyphrases
- semantic segmentation
- convolutional neural network
- street scenes
- face detection
- superpixels
- conditional random fields
- weakly supervised
- label transfer
- scene classification
- object categories
- object class
- object detection
- object classes
- neural network
- image understanding
- pascal voc
- multiscale
- pairwise
- viewpoint
- natural images
- higher order
- object recognition
- small number
- image features
- bayesian networks
- image segmentation
- feature selection
- hidden markov models