TRAJECTORY.MD

Define your camera trajectory.

1. Coordinate System Definition

In vistaDream, the coordinate system is a right-handed system where:

• Positive X points to the right,
• Positive Y points down,
• Positive Z points forward.

     / Z+
    /
   /
  +-------------> X+ 
  |
  | 
  |  
  v Y+  

2. Trajectory define

In Vistadream, the base trajectory class Traj_Base is defined in ops/trajs/basic.py.

• The trajectory is defined as a sequence of cameras (position + pose). For each camera, the following parameters need to be provided:

  • The camera position (Camera);
  • Look-at position (Target): a point along the camera's viewing direction
  • The camera's up vector (Up).

  The illustration of the above parameters are given in Fig-below(a). Moreover, As shown in Fig(b), the 'Up' direction does not necessarily align strictly with the camera's up vector. It can be any direction within the upper half of the camera's plane, as long as the angle between it and the true up direction is less than 90 degrees. This makes the 'Up' defination much simple. In general, when the camera is approximately vertical, the 'Up' vector can be simply defined as [0,-1,0].

  

• Then, You can then define your own Traj class by inheriting from this Traj_Base class. Examples are in ops/trajs/rot.py and ops/trajs/wobble.py. In your class, you only need to define a List of cameras in the camera_target_up function, where each element in the list is a tuple, defined as (Camera, Target, Up)

• Tips: Traj_Base owns a fundamental attribute that you might need to generate trajectories: radius, which approximates the radius of the scene (excluding the sky).

3. Trajectory check

If you want to verify whether the trajectory construction is correct, you can use the following code for checking, taking class Rot in ops/trajs/rot.py as a reference:

from ops.trajs.check import Base_Check_Traj
# change to yours below
from ops.trajs.rot import Rot

rgb_fn = 'data/sd_readingroom/color.png'
checkor = Base_Check_Traj(rgb_fn)
checkor._fake_scene()

# change Rot below to your class name
trajector = Rot(checkor.fake_scene,nframe=100)

checkor.trajs = trajector()
trajector._visualize_traj(checkor.trajs) # open3d visualization
checkor._render_video() # rendering check

Then you should obtain an Open3D window with trajectory visualization below and two rendering videos check_traj.video_rgb.mp4 and check_traj.video_dpt.mp4 for check.