知识卡片

普通读书笔记卡 · 9002.d · 生成模型:GPT-6

数据集身份也能被识别

理解等级:中级

Name That Dataset 游戏用少量图像让人识别来源,表明数据集带有风格、对象和构图的统计指纹。书中约 75% 是特定历史研究中熟悉数据集的研究者结果。可以训练来源分类器并用未见图像评估,作为发现来源差异的诊断;可识别不等于已证明差异造成任务损害。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      In the paper [471], the authors propose a game that we reproduce in figure 35.8. The name of the game is Name That Dataset! The figure shows three random images from 12 famous computer vision datasets and the goal of the game is to match the images with each dataset name. The point of the game is to show that it is easy to recognize

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第35章《数据偏差与分布漂移》(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/c035b.xhtml) - 结论依据:摘录中关于数据集身份也能被识别的定义、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:In the paper [471], the authors propose a game that we reproduce in figure 35.8. The name of the game is Name That Dataset! The figure shows three random images from 12 famous computer vision datasets and the goal of the game is to match the images with each dataset name. The point of the game is to show that it is easy to recognize