Weakly-supervised VisualBERT: Pre-training without Parallel Images and Captions.
Liunian Harold LiHaoxuan YouZhecan WangAlireza ZareianShih-Fu ChangKai-Wei ChangPublished in: CoRR (2020)
Keyphrases
- weakly supervised
- object detectors
- image database
- image data
- input image
- three dimensional
- image retrieval
- ground truth
- object detection
- object class
- image classification
- object recognition
- test images
- superpixels
- computer vision
- multiple images
- image annotation
- image regions
- segmentation method
- information retrieval
- co occurrence
- small number
- image segmentation