You are on page 1of 19

Drowsiness

detection
Introduction :
With the increase in traffic accidents in the world, statistics have
indicated that car accidents are the main cause of death among young
people. Therefore, it was necessary to find a solution to this human and
.material catastrophe
Solutions :
A smart system that detects the
movement of the driver's eyes and
mouth and warns him in case he closes
his eyes or yawns, and a video is filmed
to monitor the state of the face .
Similar Applications
Anti drowse
01 App 03 Anti sleep driver

Stay awake Carsafe App


02 while driving 04
Application Advantages
Detect mouth movement when
driver used sunglasses

Available Real time


Acquire video Process into Face
feed frames Detection

Project Structure: Eye


Detection
Mouth
Detection

Eye
Yawn
Closure
Detection
Detection

Sound Alarm if the


conditions are
satisfied
Face Detection: Get SVM Model
Given test images

start Load SVM model

Detection
Data set
preparation (Size
64*128)

Hog Feature
Extraction

Not
SVM Training
face
1 to positives Hog
features & 0 to
negative

Yes face
In practice this window would be
applied to multiple scales of the
image.
overlapping bounding box
problem:
o might ran into the problem of detecting
multiple bounding boxes around the
object you want to detect in the image.
Solutions :
 Applying non-maximum suppression algorithm to remove the redundant
bounding boxes.
non-maximum suppression
algorithm:
68 Facial
Landmarks
The Algorithm

01 02 03
Passes the features into an Refines the predicted locations
Examines the features to the
Ensemble of Regression Trees to improve accuracy through a
input model
(ERT) cascade of regressors
TRAINING PROCESS
Train the model
Optimize model on our optimal
parameters values
Second Fourth

First Third
extract only the Choose the most
eye and mouth appropriate values
(x,y)coordinates for our task
from the XML file
Our trained shape predictor
Capable of performing 640
predictions in a second

MAE(test-set) = 9.64
Fast
MAE(train-set) =4.29

Accurate small
The size of our model is
only 9.0 MB
EAR : Eye Aspect Ratio
●  
1. Euclidean distance:
●  

3. EAR

2.

4.
MAR mouth aspect ratio:
 
1. Euclidean distance:

2.

3.

4.
1.Track the movement

Future
of the head in addition
to the eyes and the
mouth.

prospects 2. Alerts the driver


when he exceeds the
specified speed by
developing it to be able
to check vehicle speed
3.Use an infrared camera to solve and brake speed.
the problem of detecting the mouth
and eyes in case of wearing a mask
or sunglasses.

4.Android Application
THANKS!

You might also like