![Wayve](/img/default-banner.jpg)
- Видео 144
- Просмотров 378 425
Wayve
Добавлен 2 авг 2017
Wayve is a leading developer of embodied intelligence for autonomous vehicles. We use AI to pioneer a next-generation approach to self-driving: AV2.0, which enables fleet operators to unlock the benefits of AV technology at scale.
Our solution uses deep learning to tackle the challenges of self-driving, eliminating the need for expensive and complex robotic stacks that require highly-detailed maps and programmed rules. Autonomous vehicles powered by Wayve's AI software can drive in any environment, even new places, without explicit programming.
Today, we're showcasing the promise of our technology with daily testing on UK roads. Inspired by our vision for a smarter, safer, more sustainable world, our goal is to be the first to bring autonomy to 100 cities.
Our solution uses deep learning to tackle the challenges of self-driving, eliminating the need for expensive and complex robotic stacks that require highly-detailed maps and programmed rules. Autonomous vehicles powered by Wayve's AI software can drive in any environment, even new places, without explicit programming.
Today, we're showcasing the promise of our technology with daily testing on UK roads. Inspired by our vision for a smarter, safer, more sustainable world, our goal is to be the first to bring autonomy to 100 cities.
CVPR24 E2EAI | Gianluca Corrado: Learning Models of the World
Learning Models of the World: Exploring Generative World Models in Autonomous Driving
Presented by Gianluca Corrado, Principal Scientist at Wayve. This session will focus on advancements in generative world models. Explore how integration of world models empowers autonomous vehicles to anticipate and strategize their actions, elevating safety and efficiency on the road. Discover how, by incorporating world models into driving algorithms, there exists the potential for enhanced comprehension of human decisions, ultimately facilitating better adaptability to a broader array of real-world situations.
More info: wayve.ai/cvpr-e2ead-tutorial/
Presented by Gianluca Corrado, Principal Scientist at Wayve. This session will focus on advancements in generative world models. Explore how integration of world models empowers autonomous vehicles to anticipate and strategize their actions, elevating safety and efficiency on the road. Discover how, by incorporating world models into driving algorithms, there exists the potential for enhanced comprehension of human decisions, ultimately facilitating better adaptability to a broader array of real-world situations.
More info: wayve.ai/cvpr-e2ead-tutorial/
Просмотров: 433
Видео
CVPR24 E2EAI | Nikhil Mohan: Towards Neural Simulator
Просмотров 40821 день назад
Towards Neural Simulator: Offline Validation of End-to-End Autonomous Driving Presented by Nikhil Mohan, Lead Scientist at Wayve, this session will focus on the development and application of neural simulators in autonomous driving. It will cover their roles in enhancing the validation and efficiency of end-to-end systems and their impact on advancing research in the domain. More info: wayve.ai...
CVPR24 E2EAI | Hongyang Li: Could Foundation Models really resolve End-to-end Autonomy?
Просмотров 61121 день назад
Presented by Hongyang Li, Principal Investigator at OpenDriveLab. This session will explore the evolution of autonomous driving from its early days to the current state-of-the-art technologies, focusing on the shift from modular to end-to-end approaches. More Info: wayve.ai/cvpr-e2ead-tutorial/
CVPR24 E2EAI | Jamie Shotton: Frontiers in End-to-End Learning for Autonomous Driving
Просмотров 89021 день назад
Presented by Jamie Shotton, Chief Scientist at Wayve. This segment delves into the latest advancements and methodologies in end-to-end learning tailored explicitly for autonomous driving, highlighting cutting-edge research and innovations in the field.
CVPR24 E2EAI | Long Chen: Introduction
Просмотров 87221 день назад
Presented by Long Chen, Staff Scientist at Wayve More info: wayve.ai/cvpr-e2ead-tutorial/
CVPR24 E2EAI | Oleg Sinavski: Language Meet Driving
Просмотров 36521 день назад
Language Meet Driving: Empowering End-to-End Autonomous Driving with Large Language Models (LLMs) Presented by Oleg Sinavski, Principal Scientist at Wayve. This session will focus on the usage of Large Visual-Language Models in autonomous driving. What are the benefits of adding language modality to a behaving autonomous robot? Here, we will cover topics in explainability, grounding textual mod...
CVPR24 E2EAI | Elahe Arani: Navigating the Future of End-to-End Autonomous Driving
Просмотров 37221 день назад
Navigating the Future of End-to-End Autonomous Driving: Reflections and Future Directions Presented by Elahe Arani, Head of AI Research at Wayve and Adjunct Assistant Professor in the Department of Mathematics and Computer Science at Eindhoven University of Technology. In the final session, we will reflect on the key insights gained throughout the tutorial, summarizing the fundamental concepts ...
PRISM 1 combined rgb depth
Просмотров 186Месяц назад
This section shows examples of scene reconstructions visualised alongside depth and 3D velocity magnitude rendered onto images. Note that these visualisations are reconstructions, not the real videos.
PRISM 1 combined aux vis
Просмотров 109Месяц назад
This section shows examples of scene reconstructions visualised alongside depth and 3D velocity magnitude rendered onto images. Note that these visualisations are reconstructions, not the real videos.
PRISM 1 London 4
Просмотров 150Месяц назад
PRISM-1, a scene reconstruction model of 4D scenes (3D in space time) from video data. Here we demonstrate PRISM-1’s capability to reconstruct a scene from various viewpoints by changing the camera path in two ways. When “freezing time” occurs, the ego-vehicle remains fixed in time while we pan the camera from left to right to view the scene from different angles. When the “freezing position” o...
PRISM 1 London 3
Просмотров 89Месяц назад
PRISM-1, a scene reconstruction model of 4D scenes (3D in space time) from video data. Here we demonstrate PRISM-1’s capability to reconstruct a scene from various viewpoints by changing the camera path in two ways. When “freezing time” occurs, the ego-vehicle remains fixed in time while we pan the camera from left to right to view the scene from different angles. When the “freezing position” o...
PRISM 1 Roundabout
Просмотров 68Месяц назад
PRISM-1, a scene reconstruction model of 4D scenes (3D in space time) from video data. Here we demonstrate PRISM-1’s capability to reconstruct a scene from various viewpoints by changing the camera path in two ways. When “freezing time” occurs, the ego-vehicle remains fixed in time while we pan the camera from left to right to view the scene from different angles. When the “freezing position” o...
PRISM 1 Road works
Просмотров 74Месяц назад
PRISM-1, a scene reconstruction model of 4D scenes (3D in space time) from video data. Here we demonstrate PRISM-1’s capability to reconstruct a scene from various viewpoints by changing the camera path in two ways. When “freezing time” occurs, the ego-vehicle remains fixed in time while we pan the camera from left to right to view the scene from different angles. When the “freezing position” o...
PRISM 1 Red light
Просмотров 70Месяц назад
PRISM-1, a scene reconstruction model of 4D scenes (3D in space time) from video data. Here we demonstrate PRISM-1’s capability to reconstruct a scene from various viewpoints by changing the camera path in two ways. When “freezing time” occurs, the ego-vehicle remains fixed in time while we pan the camera from left to right to view the scene from different angles. When the “freezing position” o...
PRISM 1 London 2
Просмотров 40Месяц назад
PRISM-1, a scene reconstruction model of 4D scenes (3D in space time) from video data. Here we demonstrate PRISM-1’s capability to reconstruct a scene from various viewpoints by changing the camera path in two ways. When “freezing time” occurs, the ego-vehicle remains fixed in time while we pan the camera from left to right to view the scene from different angles. When the “freezing position” o...
Ride The Wayve: Autonomous Driving in busy Central London
Просмотров 1,7 тыс.3 месяца назад
Ride The Wayve: Autonomous Driving in busy Central London
Ride the Wayve: Uninterrupted autonomous driving
Просмотров 1,4 тыс.4 месяца назад
Ride the Wayve: Uninterrupted autonomous driving
Meet Wayve's new President, Erez Dagan
Просмотров 3515 месяцев назад
Meet Wayve's new President, Erez Dagan
Ride The Wayve with Christopher Young (Full video)
Просмотров 1,4 тыс.5 месяцев назад
Ride The Wayve with Christopher Young (Full video)
Ride The Wayve with Christopher Young
Просмотров 1965 месяцев назад
Ride The Wayve with Christopher Young
Ride The Wayve: Reacting to your Road
Просмотров 4706 месяцев назад
Ride The Wayve: Reacting to your Road
Very nice indeed! to be honest you need to up the quality, of the video presentations on here.
Good that you allow comments. I think you would benefit from higher quality cameras, not just resolution but tonal and color precision. The higher quality data should allow easier learning. Also can image synthesis ever hope to be high enough quality to be useful for training? it seems like it's a fruitless approach.
This actually can replace black cab drivers , we can still have the famous cabs but without grumpy Fella’s in who drives dodgy and had cockney accent
impressive!
I wonder how long before Tesla drivers and investors start paying attention to this video. Elon, you have a problem. Btw they just gave the AI the UK highway code book as a pdf and just videos of other drivers. Wow, just wow.
Impressive!
This is certainly impressive. Once the Wayve car can meet another vehicle on a single track road, remember that it passed a passing place not too far back, and then reverse into it, then I really will bow down and say that AI is fully human equivalent, and better than most drivers.
It blows my mind that if a video like this was put out by Tesla it would have hundreds of thousands of views and hundreds of comments, and yet here is a clearly superior technology and this is the first comment, 6 months after the video was released! I was beginning to lose faith that we would see self driving in wide availability in my lifetime (as I am quite old), but this video has restored my faith. I hope Wayve prospers even though capital is no longer easy to raise.
I am quite heavily invested in Tesla stocks but I have to agree with you. The Wayve driving is natural. But with Tesla FSD videos every few minutes I am thinking, they need to work on that bit of the tech. Very impressed with Wayve.
Uber drivers soon finished.
The neural net must be wondering where all the traffic lights are ;p
What does the car do in a situation in a single lane with passing places if neither vehicle can pass without one backing up? How is that negotiated with the oncoming vehicle? Or the one behind you?
If you see a car with no driver, you back up.
Into the car behind you?
Hm yeah nothing has changed Tesla does the exact same thing and once in mass used caused many deaths this is a shame learn no one wants a true self drivng car make auto pilot not this.
Thank you. Just for curious, when regarding NeRF, is NVidia’s instant-ngp technology utilized in your work?
You should release the driving simulator to the public as a game! It would be so fun to just drive around virtually and explore. It could replace Google Street View.
What model architecture are you using here?
Five years ago this company was already experimenting with end to end. Really fascinating to see.
Your computer shows me nothing other than it doesn't give a shit about anyone on a bike and hate to tell you end to end mL isn't really new and it isn't the answer either Matter of fact that A lot of engineers stay far from it because it's a very unpredictable system and why the steering wheel all over the place
Very inspirational - great to see such visionaries marching towards the future!!
Thank you, please keep these coming, it's really excellent Wayve is London-based. Look forward to more content like this, as well as non-technical demos. So far those have been tightly controlled and cherry picked, it would be great to have much longer uncut videos.
You got your wish. Recent vids have been 30-40 mins long
Try testing in Bangladesh. Most AI cars will FAIL there
Saya tidak percaya ia boleh menjadi sebaik ini
Tesla is gonna be proud
Intuitively, this will ALMOST work....so long as you give the car a tremendous number of experiences. But I suspect it would kill a lot of people along the way, as it "learns" that sometimes the sun blinds it and sometimes people wearing camo are not landscape features, and, and, and......
This is how they do it with weaponized UAV, so it makes sense.
It's actually not the first example, OpenPilot has had End to End models since before then.
With the standard of some drivers bring on the self drive vehicles 🤖
seems like a similar approach to tesla/Openpilot, interesting
It is that’s why it’s the best approach to self driving
Soundtrack from the video: Frenchies - Moving Away 😁
the car drives mainly with semantic segmentation and depth. so the weather shouldn't matter as long as the segmentation works. what do i get wrong on this?
The depth map from a monocular view is amazing! Do you use a singel img or multiple imgs as input? motion would make the depth estimation much easyer
I hype your company so much 😅 for me as a stundet in ai .. this is like a dream! I would work for free for you. Just give me a call 😅😂
This is breathtaking. just amazing! the traffic light got me! just wow Where can I apply? ^^
this is just amazing! good job!
How I could contribute to this? I want try it on my Twizy
Maybe use a 4k camera for PR work instead of 95 webcam.
I wondered when you guys were going to move onto the Jaguar i-Pace :)
Hi, it's beautiful work. I have one question though. As the noise vector is sampled from the present distribution, how can you interpret the action name (e.g., turn left, right, straight) from the sampled vector? Let's say I want to "command" the model to imagine turn-left views. Is there any parameter that allows me to explicitly sample the "turning-left" vector from the present distribution?
Thanks! In this specific model that isn't possible, but if you condition the model on a navigation command like we did in arxiv.org/pdf/1912.00177.pdf, then you could achieve this
Amazing. I am afraid that self-driving cars will become more intelligent driver than most humans are with each passing year. They might even win in competitive car races in future like formula one.
Car is not confident enough to drive without jerking. Hopefully, it has learnt enough to become confident to drive straight and great. It is almost a year.
Amazing work
Great video guys! Keep working! Good luck!
Using asyncio, I see. People of culture.
What is different in your approach than Tesla in terms of reinforcement learning? Is small amount of RL training really enough to handle unforeseen exceptions? I am just curious not offensive. Thanks in advance. And congrats for your work.
Which RL algorithm is used? DQN, Actor- critic? Thanks in advance.
Based on their paper, it's DDPG
It's a very nice work done here. A way to go.
wow
Really Inspired from this work.
Glad to see you back, thought you had gone. How is your i Pace project going?
Have you got any videos of your self-driving cars interacting on the road with other cars? When will you do a passenger test service?
Your team make an amazing job👌