ABSTRACT

Video-based abnormal driving behaviour detection is becoming more and more popular for the time being, as it is highly important in ensuring safeties of drivers and passengers in the vehicle, and it is an essential step in realizing automatic driving at the current stage. Thanks to recent developments in deep learning techniques, this challenging detection task can be largely facilitated via the prominent generalization capability of sophisticated deep learning models as well as large volumes of video clips which are indispensable for thoroughly training these data-driven deep learning models.

deep learning fusion techniques are emphasized, and three novel deep learning-based fusion models inspired by the recently proposed and popular densely connected convolutional network (DenseNet) are introduced, to fulfill the video-based abnormal driving behavior detection task for the first time.

EXISTING SYSTEM

• abnormal driving detection and deep learning techniques, which are closely related to this study, are emphasized. Recent developments in the two aspects are briefly reviewed, with pros and cons been discussed.
• can be summarized based on literatures of automatic abnormal driving behavior detection that, there are often three commonly used detection schemes.
• The first one is based on the detection of human physiological signals (i.e., electrooculogram, electro-encephalogram, respiratory, blood flow changes, etc.) using diverse kinds of sensors [27], [28]. The second one is based on facial details [29] (i.e., changes in eye movement, mouth movement, head movement, hand features, etc.).
• deep learning techniques receive vast popularity when powerful computational hardware and large-scale data become more and more available nowadays.  
• Generally speaking, most contemporary deep learning models can be categorized into two types, i.e., deep generative learning models and deep discriminant learning models. To be specific, deep generative learning models mainly aim to replicate ‘‘fake-but-realistic’’ data based on real data, and popular deep generative learning models include but not limited to VAE (i.e., variational auto-encoder) [34], GAN (i.e., generative adversarial network) [35], GLOW (i.e., generative flow)

PROPOSED SYSTEM :-

• proposed deep learning-based fusion models in automatically detecting abnormal driving behavior of this study, the Kaggle state farm distracted driver detection database
• demonstrates the trend of accuracies increasing with respect of training epochs in all compared deep learning models. First, it can be noticed that, accuracies of all deep learning models keep on increasing and then become stable when their training epochs further increase, which is a significant indicator of the thorough training and convergence of all deep learning models. Second, three deep learning-based fusion models, Dense Net, as well as Res Net outperform other conventional CNN-based models (i.e., CNN, Wide CNN, Group CNN) as revealed in Figure 9. For comparisons between three deep learning-based fusion models and Dense Net, it is interesting to notice that, the former reaches the stable stage faster ( i.e., less epochs) than Dense Net, and significant robustness can be obtained from new deep learning-based fusion models.

ADVANTAGES OF PROPOSED SYSTEM:

• main advantage of affecting drivers’ normal driving cannot be neglected, either. Furthermore, physiological signals of human beings vary greatly due to the physiological difference in each individual person and her / his environmental conditions.

SYSTEM REQUIREMENTS
SOFTWARE REQUIREMENTS:
• Programming Language : Python
• Font End Technologies : TKInter/Web(HTML,CSS,JS)
• IDE : Jupyter/Spyder/VS Code
• Operating System : Windows 08/10

HARDWARE REQUIREMENTS:

 Processor : Core I3
 RAM Capacity : 2 GB
 Hard Disk : 250 GB
 Monitor : 15″ Color
 Mouse : 2 or 3 Button Mouse
 Key Board : Windows 08/10