Project Source Code

Instructions

1. Calibrate the two cameras (eye-to-hand) by hand_eye_calib_apriltag.py, save trajectory/extrinsics.
2. Label ground truth scan spots on the subject's body (target 1, target 2, target 4) with a mark pen.
3. Change the SUBJECT_NAME and SCAN_POSE in subject_info.py and run python subject_info.py to create a new folder for the new subject/pose.
4. SCAN_POSE is either 'front' or 'side'.
5. Have the subject lied on the stretcher.
6. cd .., and run python run_pipeline.py to run the entire pipeline, for both front and side poses.
7. Manually measure and record the pixel coordinates of the three targets in two color images.
8. For target 1 and 2, use the images taken for the front pose.
9. For target 4, use the images taken for the side pose.

Ratio Optimization

1. Run python evaluation/compute_gt_position.py to compute the ground-truth 3D positions:
2. Run python optimize.py to optimize the ratio parameters for both front and side postures.