GapFill: A Python module of gap filling functions for motion capture marker data
Collection of several functions for gap filling and recovering of motion capture marker data
GapFill can be installed from PyPI using pip on Python>=3.7.
pip install gapfillRows of ndarray values for marker coordinates should be filled with numpy.nan for occluded (blocked) frames.
import gapfill as gf
# numpy is required in order to provide necessary markers' coordinate values
import numpy as np
# 'fill_marker_gap_interp()' function will update the given ndarray by filling its gaps using bspline interpolation
# 'tgt_mkr_pos0': a 2D (n, 3) ndarray of a target marker position to fill the gaps
# 'n' is the total number of frames
tgt_mkr_pos0 = np.array((n, 3), dtype=np.float32)
# 'ret0': either True or False, True if there is any frame updated, False if there is no frame updated
# 'updated_frs_mask0': a boolean ndarray to indicate which frames are updated
ret0, updated_frs_mask0 = gf.fill_marker_gap_interp(tgt_mkr_pos0)
# 'fill_marker_gap_pattern()' function will update the given ndarray by filling its gaps using a donor marker
# 'tgt_mkr_pos1': a 2D (n, 3) ndarray of a target marker position to fill the gaps
# 'n' is the total number of frames
tgt_mkr_pos1 = np.array((n, 3), dtype=np.float32)
# 'dnr_mkr_pos': a 2D (n, 3) ndarray of a donor marker position
# 'n' is the total number of frames
dnr_mkr_pos = np.array((n, 3), dtype=np.float32)
# 'ret1': either True or False, True if there is any frame updated, False if there is no frame updated
# 'updated_frs_mask1': a boolean ndarray to indicate which frames are updated
ret1, updated_frs_mask1 = gf.fill_marker_gap_pattern(tgt_mkr_pos1, dnr_mkr_pos)
# 'fill_marker_gap_rbt()' function will update the given ndarray by filling its gaps using a cluster of 3 markers
# 'tgt_mkr_pos2': a 2D (n, 3) ndarray of a target marker position to fill the gaps
# 'n' is the total number of frames
tgt_mkr_pos2 = np.array((n, 3), dtype=np.float32)
# 'cl_mkr_pos': a 3D (m, n, 3) ndarray of the cluster markers
# 'm' (at least 3) is the number of markers, and 'n' is the total number of frames
cl_mkr_pos = np.array((m, n, 3), dtype=np.float32)
# 'ret2': either True or False, True if there is any frame updated, False if there is no frame updated
# 'updated_frs_mask2': a boolean ndarray to indicate which frames are updated
ret2, updated_frs_mask2 = gf.fill_marker_gap_rbt(tgt_mkr_pos2, cl_mkr_pos)- Smolka, J. and Lukasik, E., 2016, July. The rigid body gap filling algorithm. In 2016 9th International Conference on Human System Interactions (HSI) (pp. 337-343). IEEE.
- Arun, K.S., Huang, T.S. and Blostein, S.D., 1987. Least-squares fitting of two 3-D point sets. IEEE Transactions on pattern analysis and machine intelligence, (5), pp.698-700.
- Wikipedia: Kabsch algorithm
- Kwon3D: Computation of the Rotation Matrix
- Vicon: What Gap Filling Algorithms are used Nexus 2?
- Qualisys: Featuring the Trajectory Editor in QTM
