Run on Windows

To use the program, simply run the main.exe file.
If you prefer using a python environment, install the requirements and run the main file

pip install -r requirements.txt
python main.py

Settings

When first running the main app, a settings.json file will automatically be created in the directory where it is being run with the default settings.

Setting	Description	Value Type	Default value and examples	Nullable
video_source	You can choose which webcam to use by its index. If your computer only has one webcam, then its value is probably 0. You can also choose to use a video as input.	Int "String"	0 "path/video.mp4"	No
show_fps	Show how many frames per second are being analyzed.	Boolean	false	No
show_video	Show posture detections on the original video source (true) or on a blank canvas (false)	Boolean	true	No
mirror_mode	Inverts the camera video input horizontally.	Boolean	true	No
auto_detect_orientation	Automatically detects if the user is facing fowards or sideways. Needs to be true to have automatic alerts for bad postures	Boolean	true	No
draw_all_landmarks	Show all body landmarks that can be detected.	Boolean	false	No
draw_pose_landmarks	Show only necessary body landmarks to visualize body posture.	Boolean	true	No
vis_threshold	Minimum confidence level of a landmark to be considered as being detected, between 0 and 1.	Float	0.7	No
put_orientation_text	Show whether the user is detected to be facing fowards or sideways. The sensitivity can be adjusted with the next two settings.	Boolean	true	No
shoulder_hip_ratio_threshold	Default method of detecting orientation. Uses the ratio of shoulder-to-shoulder ditance over hip-to-hip distance. Detects facing fowards if value is over the threshold, and facing sideways if under.	Float	0.45	Yes
shoulder_height_variation_threshold	Less accurate method of detecting orientation. Uses the height difference between the two shoulders. Detects facing fowards if value is under the threshold, and facing sideways if over. Used ONLY when the previous value shoulder_hip_ratio_threshold is null.	Float	0.018	No
neck_ratio_threshold	Only when facing fowards. Uses the ratio of neck length over shoulder-to-shoulder distance. Detects bad posture when ratio is under the threshold.	Float	0.65	No
neck_angle_threshold	Only when facing sideways. Uses the smallest angle made from shoulder-to-neck and neck-to-nose angle. Detects bad posture when angle is under the threshold.	Int	60	No
resize_image_width_to	Resize image width in pixels. Preferably to be used without resize_image_height_to to keep original aspect ratio.	Int	400	Yes
resize_image_height_to	Resize image height in pixels. Preferably to be used without resize_image_width_to to keep original aspect ratio.	Int	null	Yes
time_bad_posture_alert	Alert the user only if the user is in bad posture for over certain amount of seconds.	Int	3	No
alert_sound	When alerting the user, also play a sound.	Boolean	true	No
alert_other_device	Alert the user on another device. The other device has to run this program. More info here.	Boolean	true	No
ip_address	Only when alert_other_device is true. Has to be the same IPv4 address which the second device server is listening on. More info here.	String	null "10.111.200.200"	Yes

Receiving alerts on another device

On your second device, download the following executable file server.exe, or if you prefer as a python file server.py. When running it, the IPv4 address in which the server listens to will be shown on the console. Use that address in the ip_address settings.

Creating executable file

Option 1

In the root project folder, run the following command:

pyinstaller main.spec

Option 2

pyinstaller -F main.py

then inside the main.spec file, add the following snipped under "block_cipher = None":

def get_mediapipe_path():
    import mediapipe
    mediapipe_path = mediapipe.__path__[0]
    return mediapipe_path

and add the following under "pyz = PYZ(a.pure, a.zipped_data, cipher=block_cipher)"

mediapipe_tree = Tree(get_mediapipe_path(), prefix='mediapipe', excludes=["*.pyc"])
a.datas += mediapipe_tree
a.binaries = filter(lambda x: 'mediapipe' not in x[0], a.binaries)

Now, run to create the executable file

pyinstaller main.spec

Random info

How large files are tracked using Git LFS