Distracted Driver Detection: Deep Learning vs Handcrafted Features

Hssayeni, Murtadha D.; Saxena, Sagar; Ptucha, Raymond; Savakis, Andreas

doi:10.2352/issn.2470-1173.2017.10.imawm-162

Cited by 67 publications

(36 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…is vital not only for safe driving but also for the autopilot hand-over process for the next generation self-learning AVs. Recently, there has been some progress in using CNN models in monitoring [1,16,36]. However, the adaptation of the state-of-art CNN models driven by the contextual information is yet to be explored.…”

Section: Related Workmentioning

confidence: 99%

“…Performance comparison with state-of-the-art As mentioned earlier, we are the first to report video-based activity recognition on these datasets [7,1]. The existing approaches [16,1,36] were evaluated using still images. For still images, Hssayeni et al [16] reported the accuracy of 85% using State Farm [7] and Abouelnaga et al [1] has achieved accuracy of 95.17% using their "Distracted Driver" dataset.…”

Section: M-lstm Memory Durationmentioning

confidence: 99%

“…The existing approaches [16,1,36] were evaluated using still images. For still images, Hssayeni et al [16] reported the accuracy of 85% using State Farm [7] and Abouelnaga et al [1] has achieved accuracy of 95.17% using their "Distracted Driver" dataset. However, [1] has used the validation images from the seen drivers i.e.…”

Section: M-lstm Memory Durationmentioning

confidence: 99%

“…-We are the first to report the video-based activity recognition using the State Farm dataset [7] and the "Distracted Driver" dataset [1], which are aimed to recognise driver's state/activity. All the existing approaches [16,1,36] are based on the single image classification.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Context-driven Multi-stream LSTM (M-LSTM) for Recognizing Fine-Grained Activity of Drivers

Behera

Keidel

Debnath

2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Automatic recognition of in-vehicle activities has significant impact on the next generation intelligent vehicles. In this paper, we present a novel Multi-stream Long Short-Term Memory (M-LSTM) network for recognizing driver activities. We bring together ideas from recent works on LSTMs, transfer learning for object detection and body pose by exploring the use of deep convolutional neural networks (CNN). Recent work has also shown that representations such as hand-object interactions are important cues in characterizing human activities. The proposed M-LSTM integrates these ideas under one framework, where two streams focus on appearance information with two different levels of abstractions. The other two streams analyze the contextual information involving configuration of body parts and body-object interactions. The proposed contextual descriptor is built to be semantically rich and meaningful, and even when coupled with appearance features it is turned out to be highly discriminating. We validate this on two challenging datasets consisting driver activities.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: M-lstm Memory Durationmentioning

confidence: 99%

Section: M-lstm Memory Durationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Context-driven Multi-stream LSTM (M-LSTM) for Recognizing Fine-Grained Activity of Drivers

Behera

Keidel

Debnath

2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…Feature description and extraction of these methods are handcrafted and depend on delineation of brain structures, which is arduous and prone to interand intra-rater inconsistency, or compound pre-processing of MRI images. It is also time-consuming and needs high computational cost [12,13]. Deep learning approaches, which are another family of machine learning methods, attracted many researchers working in the medical field in recent years [14].…”

Section: Introductionmentioning

confidence: 99%

Deep Learning Techniques for Automatic Detection of Embryonic Neurodevelopmental Disorders

2020

View full text Add to dashboard Cite

The increasing rates of neurodevelopmental disorders (NDs) are threatening pregnant women, parents, and clinicians caring for healthy infants and children. NDs can initially start through embryonic development due to several reasons. Up to three in 1000 pregnant women have embryos with brain defects; hence, the primitive detection of embryonic neurodevelopmental disorders (ENDs) is necessary. Related work done for embryonic ND classification is very limited and is based on conventional machine learning (ML) methods for feature extraction and classification processes. Feature extraction of these methods is handcrafted and has several drawbacks. Deep learning methods have the ability to deduce an optimum demonstration from the raw images without image enhancement, segmentation, and feature extraction processes, leading to an effective classification process. This article proposes a new framework based on deep learning methods for the detection of END. To the best of our knowledge, this is the first study that uses deep learning techniques for detecting END. The framework consists of four stages which are transfer learning, deep feature extraction, feature reduction, and classification. The framework depends on feature fusion. The results showed that the proposed framework was capable of identifying END from embryonic MRI images of various gestational ages. To verify the efficiency of the proposed framework, the results were compared with related work that used embryonic images. The performance of the proposed framework was competitive. This means that the proposed framework can be successively used for detecting END.

show abstract

Computer vision‐based recognition of driver distraction: A review

Moslemi

Soryani

Azmi

2021

Concurrency and Computation

View full text Add to dashboard Cite

Vehicle crash rates caused by distracted driving have been rising in recent years. Hence, safety while driving on roads is today a crucial concern across the world. Some of the reasons due to which drivers may lose attention include the use of mobile phones, speaking with passengers, and reaching behind to grab something while driving. There are various types of distractions, out of which we focus on manual ones based on the posture of the driver. This work presents a review on open problems and challenges concerning the vision-based distraction recognition of drivers. Types of experimental environments are described; and image modalities, public and custom datasets and body parts inspected during distraction recognition are explored. Besides, a comparative review of different works on vision-based recognition of driver's manual distractions and limitations for each dataset is presented. Main approaches of vision-based manual distractive driving recognition can be categorized into conventional and deep learning methods. These approaches are compared and classified based on whether using temporal information or not. Finally, we give some suggestions for improvement, and look forward to future development directions of the vision-based recognition of driver's manual distractions. K E Y W O R D Sdistraction recognition, driver dataset, driver gesture, driver monitoring, driving simulator INTRODUCTIONDistractions decrease driver's awareness and decision-making; lead to increased workload, delayed responses to driving events, speed and lane maintenance disruptions. 1 Distracted drivers would lose their vehicle control, resulting in crashing into vehicles or stationary objects. 2 In other words, such a driver loses vigilance regarding the driving situation, leading to a higher collision risk 3 ; then the safety of the driver and other passengers are affected. The National Highway Traffic Safety Administration (NHTSA) 4 has reported three activities, including dialing, browsing, and texting, which make drivers have their eyes off the road for a longer time, lead to threefold crash risk. Performing high cognitive load tasks while driving impacts the driver's visual behavior and driving performance. Drivers under high cognitive loads spend less time checking mirrors, traffic lights instruments, and areas around intersections. 5 This results in unintended lane drifts, overlooking other traffic participants and, in the worst scenario, crashes. 6 Rapid improvements in computers and wireless technologies have made complex and various devices that people can use while driving.Notably, in the field of entertainment, information and driver assistance, studies have proposed numerous functions. However, using such functions during driving could distract the driver from focusing on driving itself. 6 The widespread usage of cell phones, navigation, and infotainment systems has escalated the problem of distraction. Therefore, the cognitive load caused by secondary tasks has increased. Multitasking is today an essential element of ind...

show abstract

Distracted Driver Detection: Deep Learning vs Handcrafted Features

Cited by 67 publications

References 0 publications

Context-driven Multi-stream LSTM (M-LSTM) for Recognizing Fine-Grained Activity of Drivers

Context-driven Multi-stream LSTM (M-LSTM) for Recognizing Fine-Grained Activity of Drivers

Deep Learning Techniques for Automatic Detection of Embryonic Neurodevelopmental Disorders

Computer vision‐based recognition of driver distraction: A review

Contact Info

Product

Resources

About