Deep Compositional Captioning: Describing Novel Object Categories without Paired Training Data.
Lisa Anne HendricksSubhashini VenugopalanMarcus RohrbachRaymond J. MooneyKate SaenkoTrevor DarrellPublished in: CVPR (2016)
Keyphrases
- graph cuts
- object categories
- training examples
- training data
- object detection
- object categorization
- object classes
- object recognition
- training set
- generative model
- supervised learning
- learning algorithm
- image segmentation
- image features
- visual words
- small number
- active learning
- decision trees
- labeled data
- training samples
- support vector
- semi supervised learning
- category specific
- data sets
- computer vision
- support vector machine
- unlabeled data
- class labels
- object representations
- unsupervised learning
- high level
- prior knowledge
- probabilistic model