知识卡片

普通读书笔记卡 · 9093.a · 生成模型:GPT-6

实例表示还能扩展物体内部

理解等级:中级

在类别与掩码之外,可预测实例坐标、部件或形状映射,为精细理解提供额外信息。但每种输出都需定义坐标、监督与可验证含义,不能从一张二元掩码推断它已包含姿态、机械属性或可抓取点。

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

      What is interesting about this formulation is that it can be extended to provide a rich representation of the detected objects. For instance, in addition to the segmentation mask, we could also regress the u – v coordinates for each pixel using an instance-centered coordinate frame, or shape maps [165]. We can also output a segmentation of the parts.

    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:摘录中关于实例表示还能扩展物体内部的定义、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:What is interesting about this formulation is that it can be extended to provide a rich representation of the detected objects. For instance, in addition to the segmentation mask, we could also regress the u – v coordinates for each pixel using an instance-centered coordinate frame, or shape maps [165]. We can also output a segmentation of the parts.