Removing Word-Level Spurious Alignment between Images and Pseudo-Captions in Unsupervised Image Captioning.
Ukyo HondaYoshitaka UshikuAtsushi HashimotoTaro WatanabeYuji MatsumotoPublished in: EACL (2021)
Keyphrases
- input image
- image data
- image features
- image collections
- word level
- image retrieval
- preprocessing steps
- image classification
- image regions
- image analysis
- edge detection
- image content
- visual features
- language independent
- web images
- image annotation
- image segmentation
- latitude longitude
- character recognition
- semi supervised
- machine translation
- document images
- document analysis
- n gram
- handwritten documents
- machine learning