Skip to content

简单有效的工作!有一些疑问对于训练推断的数据输入 #3

@Hoantrbl

Description

@Hoantrbl

感谢作者高质量的开源,完整review完所有的code之后,仍有一些疑问向作者请教:

  1. 训练阶段,应该是只有spike作为输入信号,采取的uniform模式,去平均采样帧数生成模糊图像来生成监督信号;在测试阶段,只是采样spike的某一帧来作为监督信号,所以这个时候模型ouput的就是清晰的图,这个时候就是sharp_data作为监督?而且只有采用mid来作为测试指标的输出,请问这样的选择以及误差范围有没有考虑过呢?
  2. 在render阶段,就是直接根据插值位姿以及点云来生成对应的场景了是吗?

Metadata

Metadata

Assignees

No one assigned

    Labels

    questionFurther information is requested

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions