Weakly-supervised DCNN for RGB-D Object Recognition in Real-World Applications Which Lack Large-scale Annotated Training Data.
Li SunCheng ZhaoRustam StolkinPublished in: CoRR (2017)
Keyphrases
- weakly supervised
- real world
- training data
- object class
- weakly labeled
- superpixels
- data sets
- object detection
- relation extraction
- semi supervised
- object recognition
- supervised learning
- topic models
- decision trees
- training set
- viewpoint
- machine learning
- d objects
- prior knowledge
- learning algorithm
- object classes
- object detectors
- training samples
- object segmentation
- partial occlusion
- named entities
- color space
- training examples
- multiscale
- support vector machine
- color images
- pairwise
- co occurrence
- data mining
- range images
- segmentation algorithm
- information extraction
- image segmentation