How to render at known points

[happykk-coding]

Dear author,
I have some points, and I know their poses in the point cloud. I want to use these points as pseudo cameras to obtain images rendered from the pseudo cameras.
Currently, I have modified the render_video function in render.py to directly assign the poses of the pseudo cameras to render_poses.

FSGS/render.py

Line 63 in a536a64

render_poses = generate_ellipse_path(views)

However, the results I obtained are not quite right.

->I would like to ask, is the coordinate system of render_pose not the same as that of the point cloud? If not, what should it be?

Here is an example of my error: I selected a point on the surface of the object being held as the pose of the pseudo camera, and the rendered image I obtained is as follows. Clearly, this rendered image shows that the camera does not appear to be on the object being held.