https://github.com/sthysel/dedrowse
Dedrowse is a drowsiness monitor
https://github.com/sthysel/dedrowse
opencv python
Last synced: 2 months ago
JSON representation
Dedrowse is a drowsiness monitor
- Host: GitHub
- URL: https://github.com/sthysel/dedrowse
- Owner: sthysel
- Created: 2017-05-15T10:24:41.000Z (about 9 years ago)
- Default Branch: master
- Last Pushed: 2020-09-07T01:12:10.000Z (almost 6 years ago)
- Last Synced: 2026-01-01T07:21:57.770Z (6 months ago)
- Topics: opencv, python
- Language: Python
- Size: 68 MB
- Stars: 2
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Dedrowse is a drowsiness detector
Dedrowse is a computer vision system that can automatically detect operator
drowsiness in a real-time video stream and raise an alarm if the operator seems
to be drowsy. A operator may be a truck driver or a crane operator, or long
distance driver.
It is built on work by [Adrian Rosebrock](http://www.pyimagesearch.com/2017/05/08/drowsiness-detection-opencv/)
```zsh
Usage: dedrowse [OPTIONS]
Dedrowse drowsines detector
Options:
-p, --shape-predictor TEXT Path to facial landmark predictor [default: /ho me/thys/workspace/bhp/si/dedrowse/src/dedrowse/d ata/face.dat]
-e, --blink-ratio FLOAT Eye aspect ratio indicating blink [default: 0.3]
-t, --trigger INTEGER The number of consecutive frames the eye must be below the threshold for to set off the alarm [default: 48]
-s, --set-alarm TEXT Sound the Alarm [default: True] --alarm-sound TEXT Alarm sound file [default: /home/thys/workspace /bhp/si/dedrowse/src/dedrowse/data/alarm.wav]
-m, --alert-msg TEXT Alert message [default: DROWSINESS DETECTED]
-c, --webcam INTEGER Webcam number [default: 0]
-w, --frame-width INTEGER Width of visualization frame [default: 850]
--print-knobs Print knobs [default: False]
--help Show this message and exit.
```
It works by assuming a drowsy face's eyes are closer that normal
# Config
The follwing environmental variables can be set in the envrionment, or loaded
from a .env file.
```zsh
#DEBROWSE_ALARM=True
#DEBROWSE_AR_CONSEC_FRAMES=48
#DEBROWSE_BLINK_ASPECT_RATIO=0.3
#DEBROWSE_FRAME_WIDTH=850
#DEBROWSE_SHAPE_PREDICTOR=/home/thys/workspace/dedrowse/src/dedrowse/data/face.dat
#DEBROWSE_WEBCAM=0
#DEDROWSE_ALARM_SOUND_PATH=/home/thys/workspace/dedrowse/src/dedrowse/data/alarm.wav
#DEDROWSE_ALERT_MESSAGE=DROWSINESS DETECTED
```
# Install
Tested on a modern Arch Linux install Using Python 3.7 and OpenCV 4.0. Older
Python and OpenCV versions was also shown to work.
Check out source, enter repo and do:
```zsh
$ python -m venv ~/.virtualenvs/dedrowse
$ source ~/.virtualenvs/dedrowse/bin/activate
$ pip install -e .
```
A devpi, rpm, deb or docker image can be arranged if required.
## Old pre-OpenCV 4 Notes
This system is built arround opencv, numpy and scipy. On Arch its important to use the
pre-built Python wrappers that supports the FFMpeg video stream handling. So use the sytem
site packages as installed by pacman. Not python-opencv that ships a pre-built opencv lib in the
wheel that does not have mgeg video streaming support compiled in.
# System dependencies
Arch Linux:
```bash
$ sudo pacman -S cmake boost
```
## Development
```bash
$ mkvirtualenv --system-site-packages dedrowse
$ pip install -r requirement.txt
$ pip install -e .
```
# Resources
* Built on work from http://www.pyimagesearch.com/2017/05/08/drowsiness-detection-opencv/
* https://vision.fe.uni-lj.si/cvww2016/proceedings/papers/05.pdf
* Alarm sounds from http://soundbible.com/tags-alarm.html