知识卡片

普通读书笔记卡 · 9087.e · 生成模型:GPT-6

框精修补偿候选定位误差

理解等级:中级

分类判断候选内容,回归则调整边框以更贴近目标,两者职责不同。精修能改善合理候选,却不保证从任意背景框跳到遗漏对象;训练匹配、目标编码与允许修正幅度都会影响结果。

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

      This algorithm has four main steps. In the first step, the algorithm uses an efficient window scanning approach to produce a set of candidate bounding boxes. In the second step, we loop over all the candidate bounding boxes, we crop out the box and resize it to a canonical size, and then we apply an object classifier to classify the cropped image as containing the

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第50章《目标识别》(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/c050b.xhtml) - 结论依据:摘录中关于框精修补偿候选定位误差的定义、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:This algorithm has four main steps. In the first step, the algorithm uses an efficient window scanning approach to produce a set of candidate bounding boxes. In the second step, we loop over all the candidate bounding boxes, we crop out the box and resize it to a canonical size, and then we apply an object classifier to classify the cropped image as containing the