知识卡片
自然语言监督超越类别名词
图文配对可包含属性、动作与关系,为视觉表示提供比固定名词标签更丰富的任务信号。但描述通常只涉及图中部分内容,且可能错误或偏向显著对象;图文同现不等于每个词都被完整定位。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 视觉与语言
原书出处 · 1 处
- · 查看原书依据
The answer is yes! And it works really well. Researchers have tried using all kinds of language structures to supervise vision systems, including adjectives and verbs [231], words indicating relationships between objects [277], and question-answer pairs [17]. What seems to work best, once you reach a certain scale of data and compute, is to just use all the structure in language, and the simplest way
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第51章《视觉与语言》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c051.xhtml)
- 结论依据:摘录中关于自然语言监督超越类别名词的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:The answer is yes! And it works really well. Researchers have tried using all kinds of language structures to supervise vision systems, including adjectives and verbs [231], words indicating relationships between objects [277], and question-answer pairs [17]. What seems to work best, once you reach a certain scale of data and compute, is to just use all the structure in language, and the simplest way