Oscar: Object-Semantics Aligned Pre-training for Vision-Language Tasks.
Xiujun LiXi YinChunyuan LiPengchuan ZhangXiaowei HuLei ZhangLijuan WangHoudong HuLi DongFuru WeiYejin ChoiJianfeng GaoPublished in: CoRR (2020)
Keyphrases
- logical language
- specification language
- object identity
- complex objects
- computer vision
- definition language
- object model
- intended meaning
- d objects
- logic programming
- neural network
- logic programming language
- vision system
- programming language
- formal semantics
- natural language
- operational semantics
- language learning
- supervised learning
- action language
- conceptual graphs
- visually guided
- image processing
- real time
- highly expressive
- formal language
- denotational semantics
- proof theory
- model theoretic
- moving objects
- conceptual model
- semantically equivalent
- database query languages
- multiple inheritance
- rule language
- method invocation
- language constructs
- modal operators
- predicate calculus
- conditional logic
- speech acts
- owl ontologies
- default reasoning
- object oriented databases
- semantic information
- training set
- image sequences