Leveraging Unknown Objects to Construct Labeled-Unlabeled Meta-Relationships for Zero-Shot Object Navigation.
Yanwei ZhengChangrui LiChuanlin LanYaling LiXiao ZhangYifei ZouDongxiao YuZhipeng CaiPublished in: CoRR (2024)
Keyphrases
- spatial relationships
- d objects
- object model
- multiple objects
- data objects
- individual objects
- object identity
- moving objects
- object classes
- object segmentation
- objects represented
- similar objects
- target object
- digital objects
- bounding box
- visual objects
- object identification
- real objects
- spatial relations
- background clutter
- object level
- multi object
- supervised learning
- training data
- partially occluded objects
- object position
- object models
- object description
- object features
- training set
- complex scenes
- real world objects
- real world scenes
- labeled examples
- object representation
- object types
- deformable objects
- articulated objects
- object categories
- complex objects
- reference object
- object representations
- mobile objects
- relative position
- partial occlusion
- video sequences
- object region
- structured environments
- occluded objects
- three dimensional objects
- database objects
- rigid objects
- spatial information
- unsupervised learning
- object contours
- object orientation
- object shape
- background pixels
- curved surfaces
- object tracking
- unlabeled data
- semi supervised learning
- category labels
- active learning
- typed objects