Fatigue driving detection based on deep learning

2021 COE Engineering Design Project (XZ02)


Faculty Lab Coordinator

Xiao-Ping Zhang

Topic Category

Software Systems

Preamble

Canadian survey: 80% of traffic accidents are caused by human factors, and 10% of traffic accidents are caused by fatigue driving! Fatigue driving has long been one of the risk factors for fatal car accidents. According to the latest survey by the American Automobile Association, about 1 in 10 car accidents was caused by the driver’s fatigue. "Fatigue driving" is a big hazard, which can easily lead to traffic accidents.

Objective

Learn the biomedical knowledge, defining what is human fatigue and its expression, such as yawning, and eye closed. Detect the driver’s different fatigue status during driving. Set different fatigue levels, like level 1 – 4, which are awake, slight fatigue, medium fatigue, and heavy fatigue. Extract feature from the driver’s face, such as eyes, mouth, and whole face. Learn deep learning models for a time series.

Partial Specifications

Collect/generate video data, could be RGB or infrared data. While RGB data is useful in the daytime, but infrared data is useful at night.
Define different fatigue levels during data collection.
Select features as the input of the model.
Design a deep learning model for time series, and it should be end-to-end framework. The input is a video, and the output is the fatigue level.

Suggested Approach

For data collection, a RGB/infrared camera is fixed and used, each person could perform different levels of fatigue for training and testing.
For feature selection, you may choose the whole face, two eyes, and mouth as features, so you may first need to detect the whole face, two eyes, and mouth.
Design a deep learning model for time series data, such as Long-Short Term Memory (LSTM) etc.

Group Responsibilities

Realize and demonstrate the entire software system. Deliver the design documents for each module and interfaces.

Student A Responsibilities

Collect/generate video data, discuss with all the students.
Define different levels of fatigue, discuss with all the students.
Face detection, eye detection, mouth detection, using deep learning models, such as YOLO, work with Student B.

Student B Responsibilities

Collect/generate video data, discuss with all the students.
Define different levels of fatigue, discuss with all the students.
Face detection, eye detection, mouth detection, using deep learning models, such as YOLO, work with Student A.

Student C Responsibilities

Design deep learning models such as LSTM for fatigue detection, the result is different levels.
Turn all the parameters during training.
Collaborate with Student D.
Integrate face/eye/mouth detection and fatigue classification module together.
Coordinate with other students to make the system work, and document the system design.

Student D Responsibilities

Design deep learning models such as LSTM for fatigue detection, the result is different levels.
Turn all the parameters during training.
Collaborate with Student C.
Integrate face/eye/mouth detection and fatigue classification module together.
Coordinate with other students to make the system work, and document the system design.


Course Co-requisites

To ALL EDP Students

Due to COVID-19 pandemic, in the event University is not open for in-class/in-lab activities during the Winter term, your EDP topic specifications, requirements, implementations, and assessment methods will be adjusted by your FLCs at their discretion.

 


XZ02: Fatigue driving detection based on deep learning | Xiao-Ping Zhang | Sunday September 5th 2021 at 10:43 AM