An Empirical Evaluation of Deep Learning on Highway Driving

By Brody Huval, Tao Wang, S. Tandon, Jeff Kiske, W. Song, Joel Pazhayampallil, Mykhaylo Andriluka, Pranav Rajpurkar, Toki Migimatsu, Royce Cheng-Yue and 3 more. 2015

625 citations recorded by the source corpus.

Numerous groups have applied a variety of deep learning techniques to computer vision problems in highway perception scenarios. In this paper, we presented a number of empirical evaluations of recent deep learning advances. Computer vision, combined with deep learning, has the potential to bring about a relatively inexpensive, robust solution to autonomous driving. To prepare deep learning for industry uptake and practical applications, neural networks will require large data sets that represent all possible driving environments and scenarios. We collect a large data set of highway data and apply deep learning and computer vision algorithms to problems such as car and lane detection. We show how existing convolutional neural networks (CNNs) can be used to perform lane and vehicle detection while running at frame rates required for a real-time system. Our results lend credence to the hypothesis that deep learning holds promise for autonomous driving.

Subfields: Automotive Engineering, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Automotive Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

End-to-End People Detection in Crowded Scenes (2015, methodology): “The original version of OverFeat provided by [9] relied on an image representation trained with AlexNet [12].”
Improving Interpretability of Deep Neural Networks with Semantic Information (2017, positive): “With such success, DNNs have been integrated into various intelligent systems as a key component, e.g ., autonomous car [16, 5], medical im∗ Figure 1.”
Deep Reinforcement Learning framework for Autonomous Driving (2017, background): “This success has been replicated in lane & vehicle detection for autonomous driving [6].”
RoboNet: Large-Scale Multi-Robot Learning (2019, background): “, where data can be shared across institutions for greater levels of generalization and performance. 2Related Work Deep neural network models have been used widely in a range of robotics applications [12,13,14, 15,16,17]. However, most work in this area focuses on learning with a single robot in a s”
A Survey of Autonomous Driving: Common Practices and Emerging Technologies (2019, methodology): “Tracking in monocular images was achieved in real-time through a CNN-based method [191], [192].”
Self-Normalizing Neural Networks (2017, methodology): “Thus, when mapped across many layers, the variance in the interval [3 , 16] is mapped to a value below 3 .”
Deep UQ: Learning deep neural network surrogate models for high dimensional uncertainty quantification (2018, background): “…(NNs) (or artiﬁcial neural networks (ANNs)) are a class of function approximators that have shot to prominence in recent years because of breakthrough successes achieved in numerous artiﬁcial intelligence (AI) tasks such as image classiﬁcation [41, 42, 43, 44, 45] and autonomous driving [46, 47].”
Line-CNN: End-to-End Traffic Line Detection With Line Proposal Unit (2020, methodology): “From the results, it can be found that both the recall and precision for FCN+DBSCAN drop quickly when the IoU threshold becomes large.”
Large Language Models Are Zero-Shot Fuzzers: Fuzzing Deep-Learning Libraries via Large Language Models (2022, background): “Deep Learning (DL) is constantly providing revolutionary results and systems in critical fields like autonomous driving [34, 83], healthcare [9], and finance [67].”
Evaluating Bayesian Deep Learning Methods for Semantic Segmentation (2018, background): “With the increasing use of AI systems in real-life scenarios like autonomous driving [42, 31, 7, 40, 58] and medical diagnosis [17, 61, 56, 45], there are many cases in which additional knowledge about the model’s conﬁdence, i.e. capturing whether the model is essentially ‘guessing at random’,…”
No More Pesky Learning Rate Guessing Games (2015, background): “Deep convolutional neural networks (CNN) are giving state-of-the-art results for image recognition [13, 18, 16], object detection [7], face recognition [19], speech recognition [8], machine translation [17], image caption generation [21], and driverless car technology [10].”
MultiNet: Multi-Modal Multi-Task Learning for Autonomous Driving (2017, background): “These auxiliary tasks, such as lane or vehicle detection [13], enhance the quality of the training, resulting in improved performance on the primary task [4].”
PolyLaneNet: Lane Estimation via Deep Polynomial Regression (2020, background): “[14] were one of the first to use deep learning in lane detection.”
TensorFuzz: Debugging Neural Networks with Coverage-Guided Fuzzing (2018, background): “, 2016), in autonomous vehicles (Huval et al., 2015; Angelova et al.; Bojarski et al., 2016), as input into corporate and judicial decision making processes (Scarborough & Somers, 2006; Berk et al.”
End-to-End Deep Reinforcement Learning for Lane Keeping Assist (2016, background): “This success has been replicated in lane and vehicle detection for autonomous driving [6].”
Road Damage Detection and Classification Using Deep Neural Networks with Smartphone Images (2018, methodology): “, 2013) outperformed a previously proposed empirical method (Huval et al., 2015).”
Unsupervised Real-to-Virtual Domain Unification for End-to-End Highway Driving (2018, background): “[14] shows that convolutional neural networks can be used to do lane and vehicle detection at a real-time frame rate.”