What is the technology stack you oughta create fully autonomous vehicles?
Companies and researchers are divided on the answer to that question.
Approaches to autonomous driving range from just cameras andcomputer visionto a combination of computer vision and advanced sensors.
He also explained why Tesla is in the best position to make vision-based self-driving cars a reality.
Neural networks analyze on-car camera feeds for roads, signs, cars, obstacles, and people.
But deep learning can also make mistakes in detecting objects in images.
Lidars provided added information that can fill the gaps of the neural networks.
40% off TNW Conference!
And at test time, you are simply localizing to that map to drive around.
It is extremely difficult to create a precise mapping of every location the self-driving car will be traveling.
Its unscalable to collect, build, and maintain these high-definition lidar maps, Karpathy said.
It would be extremely difficult to keep this infrastructure up to date.
Tesla does not use lidars and high-definition maps in its self-driving stack.
And it must do all of this without having any predefined information about the roads it is navigating.
With the general vision system, you will no longer need any complementary gear on your car.
And Tesla is already moving in this direction, Karpathy says.
Previously, the companys cars used a combination of radar and cameras for self-driving.
But it has recentlystarted shipping cars without radars.
But the big question is can the synthetic neural networks do the same.
But labeling such a dataset is a great challenge.
One approach is to have it annotated manually throughdata-labeling companiesor online platforms such as Amazon Turk.
This contrasts with test-time inference, where everything happens in real-time and the deep learning models cant make recourse.
And they used radar sensor data to further verify the neural networks inferences.
All of this improved the precision of the labeling data pipe.
Karpathy did not say how much human effort was required to make the final corrections to the auto-labeling system.
But human cognition played a key role in steering the auto-labeling system in the right direction.
These included problems such as inconsistency between detection results in different cameras or between the camera and the radar.
It took four months to develop and master all these triggers.
The Tesla team went through seven iterations of data engineering.
They started with an initial dataset on which they trained their neural connection.
The errors were then revised, corrected, and if necessary, new data was added to the dataset.
We spin this loop over and over again until the online grid becomes incredibly good, Karpathy said.
Another benefit of the modular architecture of the connection is the possibility of distributed development.
Tesla is currently employing a large team of machine learning engineers working on the self-driving neural connection.
We have a team of roughly 20 people who are training neural networks full time.
Theyre all cooperating on a single neural connection, Karpathy said.
Tesla also owns and builds the AI chips installed inside its cars.
Teslas big advantage is its vertical integration.
Tesla owns the entire self-driving car stack.
It manufactures the car and the hardware for self-driving capabilities.
You get to co-design and engineer at all the layers of that stack, Karpathy said.
Theres no third party that is holding you back.
Youre fully in charge of your own destiny, which I think is incredible.
And I dont see any other company being able to reproduce Teslas approach.
Surely, object detection and velocity and range estimation play a big part in driving.
But human vision also performs many other complex functions, which scientists call the dark matter of vision.
Those are all important components in the conscious and subconscious analysis of visual input and navigation of different environments.
It will be interesting to how the technology fares against the test of time.
you might read the original articlehere.
