-
Notifications
You must be signed in to change notification settings - Fork 6
/
inverse_kinematic_guide.py
126 lines (103 loc) · 5.56 KB
/
inverse_kinematic_guide.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
#!python36
# -*- coding: utf-8 -*-
"""
NCams Toolbox
Copyright 2019-2020 Charles M Greenspon, Anton Sobinov
https://github.com/CMGreenspon/NCams
Guidelines and NCams code to run to perform an inverse kinematic analysis of the triangulated data.
Requires OpenSim 4 https://simtk.org/frs/index.php?group_id=91
Intended to be used in an interactive environment (e.g. Spyder).
Has following steps:
0. Import modules
1. Translate the triangulated data into OSim trc format
2. Scale the model
3. Run the inverse kinematic tool
4. Record the video of IKs and put it alongside the camera video
"""
# %% 0 Imports
import os
from scipy.spatial.transform import Rotation as R
import ncams
# %% 1 Translate the triangulated data into OSim trc format
BASE_DIR = os.path.join('C://', 'FLIR_cameras', 'PublicExample')
proj_path = os.path.join(BASE_DIR, '2019.12.20_8camsNoMarkers-AS-2019-12-23')
triangulated_path = os.path.join(proj_path, 'triangulated_full_rank_0.9', 'session4')
ik_dir = os.path.join(proj_path, 'inverse_kinematics')
if not os.path.isdir(ik_dir):
os.mkdir(ik_dir)
config_path = os.path.join(proj_path, 'config.yaml')
# load a csv file into a dictionary by specified column names
marker_name_dict = ncams.utils.dic_from_csv('marker_meta.csv', 'sDlcMarker', 'sOpenSimMarker')
triangulated_csv = os.path.join(triangulated_path, 'triangulated_points_4_smoothed.csv')
trc_file = os.path.join(ik_dir, 'triangulated_4_marshmallow.trc')
frame_range = (260, 360)
# makes an inverse kinematic config while importing data
ik_file = os.path.join(ik_dir, 'full_arm_model_IK_4_marshmallow.xml')
ik_out_mot_file = os.path.join(ik_dir, 'out_inv_kin_4_marshmallow.mot')
# rotate the data from the NCams coordinate system
# preview the rotations by loading the model and using 'File->Preview experimental data'
# the right click on the loaded kinematics and 'Transform'. If using our model and our
# calibration, the rotations should be as described below:
r = R.from_euler('zyx', [0, 90, 180], degrees=True)
# scipy.spatial.transform.Rotation.apply returns an ndarray with vertical vectors, so the
# function is changed in the lambda
rot = lambda v: r.apply(v)[0].tolist()
ncams.inverse_kinematics.triangulated_to_trc(
triangulated_csv, trc_file, marker_name_dict,
data_unit_convert=lambda x: x*100, # dm to mm
rate=50, zero_marker='scapula_anterior', frame_range=frame_range, rotation=rot,
ik_file=ik_file, ik_out_mot_file=ik_out_mot_file)
# %% 2 Scale the model
# Select a data subset for which you know the approximate joint angles
# In OpenSim 4 with a desired model loaded use Tools->Scale model
# Guides: https://simtk-confluence.stanford.edu/display/OpenSim/Scaling
# There are also screencasts available on youtube:
# https://www.youtube.com/user/OpenSimVideos/videos
# Manual scaling of each segment is also an option.
# %% 3 Run inverse kinematics
# In OpenSim 4 with a desired model loaded use Tools->Inverse kinematics
# Load the IK settings generated during import of data. TODO(AS)
# Guides: https://simtk-confluence.stanford.edu/display/OpenSim/Inverse+Kinematics
#
# If the thorax position was not recorded, a workaround is to unlock the thorax translation relative
# to the ground, run IK, calculate average thorax position in space and fix it there.
#
# If you are getting an error:
# InverseKinematicsTool Failed: Error reading rows in file '<FILENAME>'. Unexpected number of
# columns in line XX. Expected = 95. Received = 94.
# Thrown at trcfileadapter.cpp:186 in extendRead().
# Go to the relevant line XX (counted from the start of the file, NOT frame number) and add tab in
# the end.
# %% 4 Smooth joint angles
ik_filtered_mot_file = os.path.join(ik_dir, 'out_inv_kin_4_marshmallow_filtered.mot')
ncams.inverse_kinematics.smooth_motion(ik_out_mot_file, ik_filtered_mot_file,
median_kernel_size=11)
# %% 5 Make videos
# Load the motion generated during inverse kinematics and play it.
# To record a video, press a camera button in the top right corner of the viewer. To stop recording,
# press the button again. Save the video path to 'ik_video_path'.
video_path = os.path.join(BASE_DIR, 'exp_session_2019.12.20_videos', '4_cam19335177.mp4')
ik_video_path = os.path.join(ik_dir, '4_marshmallow.webm') # manually set filename
frame_offset = 0 # estimate manually with an external program, e.g. MPC-HC
output_path = os.path.join(ik_dir, 'marshmallow_19335177_4.mp4')
ncams.make_triangulation_video(
video_path, triangulated_csv, skeleton_config=config_path,
frame_range=frame_range, output_path=output_path,
thrd_video_path=ik_video_path,
thrd_video_frame_offset=frame_offset, # if the IK movement starts later
third_video_crop_hw=[slice(50, -100), slice(350, -700)], # crops the IK video
figure_dpi=300,
ranges=((-0.33, 3), (-2, 2), (-1.33, 6.74)), # manually set ranges for 3D plot
plot_markers=False)
# %% 6 make gifs and timeseries
# filenames = ['4_marshmallow_19335177_4', '4_pen_19335177_4']
filenames = ['4_marshmallow_19335177_4']
for filename in filenames:
# video = os.path.join(ik_dir, '{}.mp4'.format(filename))
# video_images_dir = os.path.join(ik_dir, '{}'.format(filename))
# ncams.image_tools.video_to_images(video, output_directory=ik_dir, output_format='jpeg')
# video_images = ncams.utils.get_image_list(path=video_images_dir)
# gif = os.path.join(ik_dir, '{}.gif'.format(filename))
# ncams.image_tools.images_to_video(video_images, gif, fps=25)
video = os.path.join(ik_dir, '{}_vertical.mp4'.format(filename))
ncams.image_tools.video_to_timeseries(video, num_images=5, figure_size=(9, 5), figure_dpi=300)