← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 9087 · 生成模型:GPT-6

检测输出是一组带类位置

理解等级:中级

边界框检测为各实例输出类别分数和位置,可用两角坐标或中心宽高表示。训练、回归、IoU与可视化需采用同一约定,且真实对象数量可变;边框是图像空间位置,不直接提供三维距离与姿态。

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

      The class vector, ŷi, has the same structure as in the image classification task but now it is applied to describe each bounding box. Bounding boxes are usually represented as a vector of length 4 using the coordinates of the two corners, b = [x1, y1, x2, y2], or with the center coordinates, width, and height, b = [xc, yc, w, h]. Our goal is

    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:摘录中关于检测输出是一组带类位置的定义、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 class vector, ŷi, has the same structure as in the image classification task but now it is applied to describe each bounding box. Bounding boxes are usually represented as a vector of length 4 using the coordinates of the two corners, b = [x1, y1, x2, y2], or with the center coordinates, width, and height, b = [xc, yc, w, h]. Our goal is