Example¶
The EdiHeadyTrack sample videos used here can be downloaded directly from the Github or in the terminal using wget.
Installing EdiHeadyTrack:
[ ]:
!pip install EdiHeadyTrack
Importing the package (once installation is complete):
[2]:
import EdiHeadyTrack as eht
Navigate to the directory from which video files will be accessed:
[ ]:
import os
os.chdir('..')
print(os.getcwd())
Import video to perform tracking on and create a camera object:
[4]:
video = eht.Video(filename='resources/high_speed_videos/theta_0_v_43.mp4')
camera = eht.Camera()
Perform head pose estimation on chosen video using chosen PoseDetectors from ‘posedetector’ module:
[5]:
mediapipe = eht.MediaPipe(video=video, camera=camera)
tddfa = eht.TDDFA_V2(video=video, camera=camera)
INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
------------------------------------------------------------------------------------------------------------------------
Creating MediaPipe object for video resources/high_speed_videos/theta_0_v_43.mp4: 16:01:01
------------------------------------------------------------------------------------------------------------------------
------------------------------------------------------------
Video data:
------------------------------------------------------------
Video resolution: 1280 x 720 pixels
Length of video: 1178 frames
Framerate: 1200 fps
------------------------------------------------------------
Running MediaPipe Face Mesh on selected video...
100%|██████████| 1178/1178 [01:42<00:00, 11.46it/s]
Face tracking complete...
------------------------------------------------------------------------------------------------------------------------
MediaPipe object complete! 16:02:45
------------------------------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------------------------------
Creating TDDFA_v2 object for video resources/high_speed_videos/theta_0_v_43.mp4: 16:02:45
------------------------------------------------------------------------------------------------------------------------
------------------------------------------------------------
Video data:
------------------------------------------------------------
Video resolution: 1280 x 720 pixels
Length of video: 1178 frames
Framerate: 1200 fps
------------------------------------------------------------
100%|██████████| 1178/1178 [01:20<00:00, 14.54it/s]
Face tracking complete...
Dump to TDDFA_tracking.mp4
------------------------------------------------------------------------------------------------------------------------
3DDFA_v2 object complete! 16:04:12
------------------------------------------------------------------------------------------------------------------------
Create filters to be applied to signals:
[6]:
filter = eht.Filter().low_pass_butterworth(fs=24000, lowcut=200, order=4)
filter_wax9 = eht.Filter().low_pass_butterworth(fs=2000, lowcut=200, order=4)
Create a Head object and apply the above filter:
[7]:
head_MP = eht.Head(posedetector=mediapipe, id='MP').apply_filter(filter)
head_TDDFA = eht.Head(posedetector=tddfa, id='TDDFA').apply_filter(filter)
------------------------------------------------------------------------------------------------------------------------
Creating Head object for MediaPipe Face Detector with video resources/high_speed_videos/theta_0_v_43.mp4: 16:04:12
------------------------------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------------------------------
Head object complete! 16:04:12
------------------------------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------------------------------
Creating Head object for <EdiHeadyTrack.posedetector.TDDFA_V2 object at 0x7f8a3142a040>: 16:04:13
------------------------------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------------------------------
Head object complete! 16:04:13
------------------------------------------------------------------------------------------------------------------------
Load data from a chosen IMU using a class from the imu module:
[8]:
wax9 = eht.Wax9(filename='resources/high_speed_wax9/theta_0_v_43.csv', time_offset=-7.82, id='WAX-9').apply_filter(filter_wax9)
Create a list of key times for showing video frames, and create a comparison plot:
[11]:
key_times = [0.1,0.2,0.3,0.4,0.5,0.6]
plot = eht.Plot(wax9, head_MP, head_TDDFA).plot_property('velocity', xlim=(0.1,0.6), ylim=(-250,250), key_times=key_times)
The Kernel crashed while executing code in the the current cell or a previous cell. Please review the code in the cell(s) to identify a possible cause of the failure. Click <a href='https://aka.ms/vscodeJupyterKernelCrash'>here</a> for more info. View Jupyter <a href='command:jupyter.viewOutput'>log</a> for further details.
Produce a table summarising the maximum values of each plot:
[10]:
plot.summarise()
----------------------------------------------
WAX-9 plot summary
----------------------------------------------
yaw, velocity 72.14
pitch, velocity 200.48
roll, velocity 64.86
----------------------------------------------
MP plot summary
----------------------------------------------
yaw, velocity 101.28
pitch, velocity 150.1
roll, velocity 49.88
----------------------------------------------
TDDFA plot summary
----------------------------------------------
yaw, velocity 87.29
pitch, velocity 177.25
roll, velocity 60.99
----------------------------------------------