Hi fellows,
I have got a question about the coordinate system for the synthetic data. Specifically, I want to know the transformation from the GT's coordinate to the image coordinate for each sequence.
To better illustrate, for a vSLAM problem, we usually set the first frame as the origin point (0, 0, 0) with x pointing to the image width, y pointing to the height, and z pointing forward orthogonally to the image plane.
However, I found some problems when I am profiling the data with a simple odometry system. Let's take sequence 01 for example. Following the image coordinate, the motion at the beginning of the sequence should be going directly in the Z direction. In comparison, the GT trajectory indicates movements in the Y direction.
So, to better align the trajectory between them, I would need a more detailed definition of the coordinate system.
Thanks for your attention on this.
Created by DJNing Hi @DJNing,
have you seen this [here](https://www.synapse.org/#!Synapse:syn29430445)? I used this script to evaluate the z-direction of the first two images in Sequence 1 of the first dataset. It should be positive!
```
poses_quat, poses_mat = get_poses('1', '/rootfolder/')
frame1 = poses_mat[0,:,:]
frame2 = poses_mat[1,:,:]
relative_pose = get_relative_pose(frame1, frame2)
print(relative_pose[2,-1]>0)
```
Does that make sense?
Drop files to upload
Questions regarding the coordinate system page is loading…