DSTA: Reinforcing Vision-Language Understanding for Scene-Text VQA with Dual-Stream Training Approach
IEEE SIGNAL PROCESSING LETTERS（2025）
关键词
Optical character recognition,Training,Visualization,Feature extraction,Transformers,Text recognition,Contrastive learning,Belts,Predictive models,Decoding,Visual question-answering,multimodal information,sence-text understanding
AI 理解论文
溯源树
样例
生成溯源树，研究论文发展脉络
Chat Paper
正在生成论文摘要