Oscar: Object-Semantics Aligned Pre-training for Vision-Language Tasks.
Xiujun LiXi YinChunyuan LiPengchuan ZhangXiaowei HuLei ZhangLijuan WangHoudong HuLi DongFuru WeiYejin ChoiJianfeng GaoPublished in: ECCV (30) (2020)
Keyphrases
- logical language
- operational semantics
- formal language
- specification language
- highly expressive
- visually guided
- computer vision
- formal semantics
- object identity
- dynamic logic
- object model
- programming language
- model theoretic
- intended meaning
- definition language
- natural language
- logic programming language
- conditional logic
- image processing
- denotational semantics
- semantically equivalent
- multiple inheritance
- training process
- training set
- proof theory
- moving objects
- language constructs
- description logics
- d objects
- natural language processing
- object detectors
- conceptual graphs
- complex objects
- first order logic
- logic programming