← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 9047.c · 生成模型:GPT-6

换视点检查三维重建

理解等级:中级

把预测深度反投影为点云,从不同视点观察,可暴露原视角彩色图不易显出的形状弯曲、断层和漂浮点。若可视化先去掉深度边界点,应公开该过滤,不能用过滤后外观代表完整预测。定量与独立几何参照仍不可替代。

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

      Just looking at the depth map from figure 43.5(b) might seem like the estimates are good. But it is hard to tell how good the estimation is with this visualization. The best way of appreciating the quality of a 3D reconstruction is by translating the depth map into a point cloud and then reconstructing the scene under different view points. Figure 43.6 shows two different

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第43章《单图深度估计》(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/c043.xhtml) - 结论依据:摘录中关于换视点检查三维重建的定义、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:Just looking at the depth map from figure 43.5(b) might seem like the estimates are good. But it is hard to tell how good the estimation is with this visualization. The best way of appreciating the quality of a 3D reconstruction is by translating the depth map into a point cloud and then reconstructing the scene under different view points. Figure 43.6 shows two different