What question did this study set out to answer?

To train a convolutional neural network that maps raw visual input from a single front-facing camera directly to steering commands without explicit decomposition into intermediate tasks.

April 25, 2016Open Access

End to End Learning for Self-Driving Cars

View Full Paper

Ask AI

Bookmark

To train a convolutional neural network that maps raw visual input from a single front-facing camera directly to steering commands without explicit decomposition into intermediate tasks.
Trained a convolutional neural network end-to-end using human steering angles as the supervisory signal mapped from raw single-camera video frames.
Implemented model training on an NVIDIA DevBox and executed onboard real-time inference using an NVIDIA DRIVE PX system running Torch 7 at 30 frames per second.
The vehicle autonomously navigated traffic on highways, marked and unmarked local roads, unpaved surfaces, and parking lots using minimal human training data.
The network autonomously learned internal representations for identifying road edges and guidance features without explicit human-defined intermediate objectives, operating reliably at 30 frames per second.

Authors

MBMariusz BojarskiSupélec DTDavide TestaUniversity of Pisa DDDaniel DworakowskiUniversity of Toronto

Discussion

No takes yet. Share an insight, caveat, or question.

Implication

Experimental study demonstrates direct mapping of raw camera pixels to steering commands in autonomous vehicles, suggesting end-to-end neural networks outperform modular driving architectures.

Cite This Study

Bojarski et al. (2016) studied this question.

synapsesocial.com/papers/6a0ec678b7cc3b883f22b9dc https://doi.org/10.48550/arxiv.1604.07316

View Full Paper

Ask AI

Bookmark

Implication

Experimental study demonstrates direct mapping of raw camera pixels to steering commands in autonomous vehicles, suggesting end-to-end neural networks outperform modular driving architectures.

End to End Learning for Self-Driving Cars

Key Points

Discussion

Implication

Cite This Study

End to End Learning for Self-Driving Cars

Key Points

Discussion

Implication

Cite This Study