How should I use this project to complete the pose estimation of the target object？ about lightglue HOT 2 CLOSED

Hi @GhYang0519

Yes, you can use this project to obtain better correspondences for PnP. You would need to build a sparse SfM model for the target object (e.g. from a set of images), and then you can estimate the camera pose of your RGB image in the target object frame (which you can transform to obtain the pose matrix in the camera coordinate system).

You can run all these steps with hloc, LightGlue is already integrated there. Checkout the demo notebook (also available in colab) to get started.

from lightglue.

thank you for your reply
Can I get the corresponding points between the current RGB image and the template RGB image through LightGlue, and calculate homography transformation through these corresponding points, and then for the template RGB image, I already know the 2D-3D correspondence of the target object in the image, At this point, can I calculate the pose matrix of the target object in the current RGB image?

from lightglue.