Video wird geladen...

Video konnte nicht geladen werden

Beim Laden dieses Videos ist ein Problem aufgetreten. Dies könnte an einem vorübergehenden Netzwerkproblem liegen oder das Video ist möglicherweise nicht verfügbar.

Lighting differences can make a huge difference in robotics. Today, I found a quirk in my model exemplifying this. > I collected 10h of training data. > 3h in, I notice that the left arm following the right arm for the final movement could be good for the final... show more

Dominique Paul

7,336 subscribers

97,897 Aufrufe • vor 2 Monaten •via X (Twitter)

Bildung Wissenschaft & Technologie

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

World Labs CEO Dr. Fei-Fei Li & SceniX Co-Founder Yunzhu Li on the data bottleneck in robotics and how world models help: "The lack of data in training, the lack of data in evaluation, this is very, very different from language models, where data is abundant on the internet." "And we know that in order for robotics to work, we have to somehow unlock the power of scaling law. But where does that come from?" "It's a profound problem that everybody's battling with in robotics." "We see a lot of unlock in being able to do this whole process through the modeling of the environments." "Being able to create these digital worlds... that's just going to unlock so much more potential for being able to replace all the costly and the unsafe data in the real environments with the data generated from the worlds for robots to be able to do scalable learning and evaluations." Fei-Fei Li Yunzhu Li martin_casado

World Labs CEO Dr. Fei-Fei Li & SceniX Co-Founder Yunzhu Li on the data bottleneck in robotics and how world models help: "The lack of data in training, the lack of data in evaluation, this is very, very different from language models, where data is abundant on the internet." "And we know that in order for robotics to work, we have to somehow unlock the power of scaling law. But where does that come from?" "It's a profound problem that everybody's battling with in robotics." "We see a lot of unlock in being able to do this whole process through the modeling of the environments." "Being able to create these digital worlds... that's just going to unlock so much more potential for being able to replace all the costly and the unsafe data in the real environments with the data generated from the worlds for robots to be able to do scalable learning and evaluations." Fei-Fei Li Yunzhu Li martin_casado

a16z

72,911 Aufrufe • vor 15 Tagen

I was hitless in Dark Souls 2 using a saxophone all the way to the final boss for the first time, on the cusp of a World's First. ... and then I forgot that there's a bottomless pit in the room.

I was hitless in Dark Souls 2 using a saxophone all the way to the final boss for the first time, on the cusp of a World's First. ... and then I forgot that there's a bottomless pit in the room.

Dr. Doot 🎷

783,982 Aufrufe • vor 1 Jahr

Genuinely didn't feel like finishing this due to lots of different things, but after seeing how many people wanted to see the final, it motivated me to make the final render. I still think it could be better, but here's the final. (Credits in comment below).

Genuinely didn't feel like finishing this due to lots of different things, but after seeing how many people wanted to see the final, it motivated me to make the final render. I still think it could be better, but here's the final. (Credits in comment below).

ElianoAnimates

762,420 Aufrufe • vor 1 Jahr

Marta Kostyuk after beating Tatjana Maria to reach Rouen final “Another final this year. You had a first final in Brisbane. What did you learn that can help for tomorrow’s final?” Marta: “Well for sure it helps that I don’t play #1 in the world again in the final. 😂 I lost my last 3 finals.. So I’m very excited for tomorrow. I hope for your support guys. I’ll need it tomorrow. It’s always great to be in the final regardless of how it will go. Very happy to be in the final.” ❤️

Marta Kostyuk after beating Tatjana Maria to reach Rouen final “Another final this year. You had a first final in Brisbane. What did you learn that can help for tomorrow’s final?” Marta: “Well for sure it helps that I don’t play #1 in the world again in the final. 😂 I lost my last 3 finals.. So I’m very excited for tomorrow. I hope for your support guys. I’ll need it tomorrow. It’s always great to be in the final regardless of how it will go. Very happy to be in the final.” ❤️

The Tennis Letter

39,566 Aufrufe • vor 3 Monaten

NEC Director Kevin Hassett on whether POTUS has taken him out of the running for Fed Chairman: HASSETT: "From the beginning, he and I have talked about whether it's better for me to be here in the West Wing or over at the Fed. I don't think he's made a final call on that, but I gotta to say that I'm even humbled still, perhaps, to be in the running...There are a lot of great candidates and it could well be that the president's right to make the decision that this is the best place for me right now."

NEC Director Kevin Hassett on whether POTUS has taken him out of the running for Fed Chairman: HASSETT: "From the beginning, he and I have talked about whether it's better for me to be here in the West Wing or over at the Fed. I don't think he's made a final call on that, but I gotta to say that I'm even humbled still, perhaps, to be in the running...There are a lot of great candidates and it could well be that the president's right to make the decision that this is the best place for me right now."

Jacqui Heinrich

34,589 Aufrufe • vor 6 Monaten

This looks like another step toward removing the data bottleneck in robotics. The bottleneck exists because recording demonstrations is easy, but labeling every grasp, hold, move, and release is slow and expensive. Perceptron Egocentric dropped a new robotics video annotation system. You give it raw robot-camera or first-person/egocentric video. It gives back structured labels that look more like training data: where each small manipulation starts and ends, what the subtask is, what each hand is doing, where the hands are frame by frame, and which hand is left or right. So, Input: raw video of a person or robot doing a task. Output: a machine-readable breakdown of the physical actions inside that video. Basically, it is a machine that converts robot video into policy-training supervision. That means timestamps, subtask boundaries, per-hand actions, and left-right hand grounding. So this will turn large piles of robot or egocentric video into supervision for training robot policies.

This looks like another step toward removing the data bottleneck in robotics. The bottleneck exists because recording demonstrations is easy, but labeling every grasp, hold, move, and release is slow and expensive. Perceptron Egocentric dropped a new robotics video annotation system. You give it raw robot-camera or first-person/egocentric video. It gives back structured labels that look more like training data: where each small manipulation starts and ends, what the subtask is, what each hand is doing, where the hands are frame by frame, and which hand is left or right. So, Input: raw video of a person or robot doing a task. Output: a machine-readable breakdown of the physical actions inside that video. Basically, it is a machine that converts robot video into policy-training supervision. That means timestamps, subtask boundaries, per-hand actions, and left-right hand grounding. So this will turn large piles of robot or egocentric video into supervision for training robot policies.

Rohan Paul

18,903 Aufrufe • vor 1 Monat

The next “I wish I bought Bitcoin in 2013” moment might be robotics IPOs, according to EllioTrades. "Robotics are going to represent this unique opportunity in our lifetimes to bet on something which could fundamentally change humanity."

The next “I wish I bought Bitcoin in 2013” moment might be robotics IPOs, according to EllioTrades. "Robotics are going to represent this unique opportunity in our lifetimes to bet on something which could fundamentally change humanity."

zynn

42,903 Aufrufe • vor 6 Monaten

Below is our open-source arm assembled by one of the discord members! Also, today Physical Intelligence shared their incredible top-notch robotics model Pi0. And I just want to remind you that you can use our arm together with this model and with LeRobot The arm has the same DOF as the arm used by Pi0, but has smaller size and cost starts from 200$ per arm. You can use better motors like we did in our demos, then it will be more precise and stable but the price will be about $600 per arm. Printable files, purchase list, and help with the assembly - in our discord channel.

Below is our open-source arm assembled by one of the discord members! Also, today Physical Intelligence shared their incredible top-notch robotics model Pi0. And I just want to remind you that you can use our arm together with this model and with LeRobot The arm has the same DOF as the arm used by Pi0, but has smaller size and cost starts from 200$ per arm. You can use better motors like we did in our demos, then it will be more precise and stable but the price will be about $600 per arm. Printable files, purchase list, and help with the assembly - in our discord channel.

Igor Kulakov

53,711 Aufrufe • vor 1 Jahr

Here is a fairly detailed video on why I didn't like the drill posted my my pal Arnie McNair 🇺🇸 on his account yesterday. As I say here, any drill might be good for somebody and the one in question could work for the right golfer. I also include a drill I think would be better for most.

Here is a fairly detailed video on why I didn't like the drill posted my my pal Arnie McNair 🇺🇸 on his account yesterday. As I say here, any drill might be good for somebody and the one in question could work for the right golfer. I also include a drill I think would be better for most.

Brian Manzella

14,667 Aufrufe • vor 5 Monaten

I wrote some new code to drive X-Plane directly rather than using the internal FDR playback, this lets me add in the Helicopter. Given the lack of precise data on the helicopter's route I made a rough estimate on speed and altitude. And you can see it's hidden against city lights for a long time. Be aware that the cockpit model is likely to be different and there's no data being sent to the gauges.

I wrote some new code to drive X-Plane directly rather than using the internal FDR playback, this lets me add in the Helicopter. Given the lack of precise data on the helicopter's route I made a rough estimate on speed and altitude. And you can see it's hidden against city lights for a long time. Be aware that the cockpit model is likely to be different and there's no data being sent to the gauges.

Scott Manley

144,051 Aufrufe • vor 1 Jahr

In another life I would have been a lighting designer for major concerts. And this would be the golden age. The number of unique lighting elements you can deliver per watt today is… 🤯

In another life I would have been a lighting designer for major concerts. And this would be the golden age. The number of unique lighting elements you can deliver per watt today is… 🤯

Matthew Prince 🌥

35,877 Aufrufe • vor 1 Jahr

Here's some of the DLSS 5 material we saw in the demos but didn't get a chance to film. Here I think you can see the strengths of DLSS 5 - reflections become much more attractive. Starfield doesn't have great lighting to begin with, so the differences can be profound.

Here's some of the DLSS 5 material we saw in the demos but didn't get a chance to film. Here I think you can see the strengths of DLSS 5 - reflections become much more attractive. Starfield doesn't have great lighting to begin with, so the differences can be profound.

Oliver Mackenzie

172,172 Aufrufe • vor 4 Monaten

Elina Svitolina on facing Coco Gauff next in Rome final “It’s amazing.. the feeling is just unreal to be after so many years here again in the final. It’s such an amazing feeling to do it in such a great way I think.” ❤️ “I definitely have a game plan. Not now exactly, but I will have it. We’ve played many times. It’s not a surprise. We both know the way we bring the fight. Our last match was something out of this world. I’m gonna rewatch that to find the ways.. some tactics. I want to enjoy this win tonight. Tomorrow it’s a day off.. so I can rest, prepare, and be ready for the final.”

Elina Svitolina on facing Coco Gauff next in Rome final “It’s amazing.. the feeling is just unreal to be after so many years here again in the final. It’s such an amazing feeling to do it in such a great way I think.” ❤️ “I definitely have a game plan. Not now exactly, but I will have it. We’ve played many times. It’s not a surprise. We both know the way we bring the fight. Our last match was something out of this world. I’m gonna rewatch that to find the ways.. some tactics. I want to enjoy this win tonight. Tomorrow it’s a day off.. so I can rest, prepare, and be ready for the final.”

The Tennis Letter

51,198 Aufrufe • vor 3 Monaten

#WATCH | Chittorgarh, Rajasthan: On securing first rank in UPSC civil services final result 2025, Anuj Agnihotri says, "I'm feeling good. I gave my first attempt for the UPSC CSE in 2023. I started preparing during my internship in 2022, and it's 2026 now, so obviously, it's been a long time. The journey to UPSC requires a lot of effort. It's a gruelling journey, but the final result is good, so everything feels good..."

#WATCH | Chittorgarh, Rajasthan: On securing first rank in UPSC civil services final result 2025, Anuj Agnihotri says, "I'm feeling good. I gave my first attempt for the UPSC CSE in 2023. I started preparing during my internship in 2022, and it's 2026 now, so obviously, it's been a long time. The journey to UPSC requires a lot of effort. It's a gruelling journey, but the final result is good, so everything feels good..."

ANI

251,166 Aufrufe • vor 5 Monaten

The Signs 🪧 . This is actually the first signs play of the year that I hate. This could be a good thing. I could see this hitting in the first quarter.

The Signs 🪧 . This is actually the first signs play of the year that I hate. This could be a good thing. I could see this hitting in the first quarter.

Jersey Jerry

283,724 Aufrufe • vor 8 Monaten

🗑 #Live3D #Live2D #Live2DWIP 🗑 When building a Live2D-based 2D Pseudo-3D model, the way the ArtMesh vertices are connected can also have a noticeable impact on the deformation result. Using this arm twist as an example, whether the ArtMesh edge / diagonal flow leans left or right becomes a key factor in whether jagged edges appear after twisting. When raising an arm, the human arm joint can almost only perform a 180° twist with the thumb facing toward the body. So I chose to make all the edges flow to the right, in order to improve the smoothness of the arm after twisting. This does mean that when performing a 180° twist with the thumb facing toward the camera, jagged edges will appear along the arm. But since human arms rarely rotate that way in normal use, I think this is a reasonable trade-off based on the actual use case. On the other hand, the subdivision density of the Warp Deformer also affects the twisting result. For a 180° twist, the subdivision should ideally be 18 or higher. For a 90° twist, the subdivision should ideally be 9 or higher. If you force the deformation beyond its limit, it will twist into something like thorns.😈

🗑 #Live3D #Live2D #Live2DWIP 🗑 When building a Live2D-based 2D Pseudo-3D model, the way the ArtMesh vertices are connected can also have a noticeable impact on the deformation result. Using this arm twist as an example, whether the ArtMesh edge / diagonal flow leans left or right becomes a key factor in whether jagged edges appear after twisting. When raising an arm, the human arm joint can almost only perform a 180° twist with the thumb facing toward the body. So I chose to make all the edges flow to the right, in order to improve the smoothness of the arm after twisting. This does mean that when performing a 180° twist with the thumb facing toward the camera, jagged edges will appear along the arm. But since human arms rarely rotate that way in normal use, I think this is a reasonable trade-off based on the actual use case. On the other hand, the subdivision density of the Warp Deformer also affects the twisting result. For a 180° twist, the subdivision should ideally be 18 or higher. For a 90° twist, the subdivision should ideally be 9 or higher. If you force the deformation beyond its limit, it will twist into something like thorns.😈

📐Hephaestus📏Live2D匠人魂

11,144 Aufrufe • vor 1 Monat

Full interview by Yuki Ishikawa for tbs before leaving for the Philippines: (His words 🥹) "I think we're a team that gets a lot of support, so I hope we can use that support to achieve good results. "I think we are a team that gets a lot of support, so I hope we can use this support to get good results. I think there will be a lot of people there; we'll be in the Philippines, so the atmosphere will be in our favor. I'm not particularly worried about that. We had a very disappointing result in the VNL, and if we manage to advance well in this WCH, we'll face Poland in the quarterfinals. I think it's another big challenge, but first we want to win one game at a time in the qualifiers. Of course, our goal is to reach the podium, but first, we have to compete in the qualifiers, so I think the first match is the most important, so I want to be especially aware of that." fighting 💪🏻

Full interview by Yuki Ishikawa for tbs before leaving for the Philippines: (His words 🥹) "I think we're a team that gets a lot of support, so I hope we can use that support to achieve good results. "I think we are a team that gets a lot of support, so I hope we can use this support to get good results. I think there will be a lot of people there; we'll be in the Philippines, so the atmosphere will be in our favor. I'm not particularly worried about that. We had a very disappointing result in the VNL, and if we manage to advance well in this WCH, we'll face Poland in the quarterfinals. I think it's another big challenge, but first we want to win one game at a time in the qualifiers. Of course, our goal is to reach the podium, but first, we have to compete in the qualifiers, so I think the first match is the most important, so I want to be especially aware of that." fighting 💪🏻

joey, and flower 🌼

94,088 Aufrufe • vor 11 Monaten

Might of found a way to use this mechanic in comp. I've seen people do this on the floor a couple times but not really on the wall in this way. I do think this could be used in comp because of the sharp change in trajectory of the ball and it could seriously bait out players. Not to mention it can be done on any part of the wall so even off the ceiling. I'm already getting fairly consistent at it on one side and I've practiced it for a day. It feels like doing the breezi flick for the first time but this could be a more practical mechanic. If it can be done on command consistently it has ridiculous potential for passes and other stuff Not saying this is a new mechanic btw nor is it guaranteed to become meta but I love the idea of seeing innovation and new things in the game because there isn't much of it these days so I hope people aren't closed minded about it. Would love to hear opinions on it

Might of found a way to use this mechanic in comp. I've seen people do this on the floor a couple times but not really on the wall in this way. I do think this could be used in comp because of the sharp change in trajectory of the ball and it could seriously bait out players. Not to mention it can be done on any part of the wall so even off the ceiling. I'm already getting fairly consistent at it on one side and I've practiced it for a day. It feels like doing the breezi flick for the first time but this could be a more practical mechanic. If it can be done on command consistently it has ridiculous potential for passes and other stuff Not saying this is a new mechanic btw nor is it guaranteed to become meta but I love the idea of seeing innovation and new things in the game because there isn't much of it these days so I hope people aren't closed minded about it. Would love to hear opinions on it

Breezi

372,865 Aufrufe • vor 2 Jahren

Experiments in progress. The one on the right has been learning for ~3 hours, the one in the middle for ~1 hour, and the one on the left just started a few minutes ago. The initial motivation for making the physical Atari was just to commit ourselves to a subset of algorithms that can make progress in this setup. This commitment rules out algorithms that require billions of samples to learn (or worse, require multiple environments running in parallel). Atari games are simple enough that we should be able to show learning on them in a short amount of time with no prior knowledge. Since then, I've realized that this setup is also a good way to compare different paradigms in robotics in a principled way. These paradigms are sim2real, learning from tele-operated data, and learning directly on the robots. So far, I have observed that getting sim2real to work reliably is hard. It requires tweaks that don't scale. Policies that can play perfectly in simulation fall apart because of latencies and the messiness of the real world. These aspects could be modeled to improve the simulation, but not without sinking significant human engineering hours. I have higher hopes for learning from tele-operated data, but that requires a human to learn the task first. These experiments are on my to-do list. I have to learn to play some of the games well through the robot. I’m half-decent at playing Pong and Ms Pacman now. Learning directly on robots is looking like the most promising approach. This approach takes away pesky distribution shifts and makes it possible to have algorithms that continually improve with more data and time without any human intervention. It feels great to let experiments run overnight and wake up to find improved policies. With learning on robots, I should, in principle, be able to go on a long vacation and come back to find better policies for complex tasks beyond Atari games. Whether that is possible with current learning algorithms is a different question.

Experiments in progress. The one on the right has been learning for ~3 hours, the one in the middle for ~1 hour, and the one on the left just started a few minutes ago. The initial motivation for making the physical Atari was just to commit ourselves to a subset of algorithms that can make progress in this setup. This commitment rules out algorithms that require billions of samples to learn (or worse, require multiple environments running in parallel). Atari games are simple enough that we should be able to show learning on them in a short amount of time with no prior knowledge. Since then, I've realized that this setup is also a good way to compare different paradigms in robotics in a principled way. These paradigms are sim2real, learning from tele-operated data, and learning directly on the robots. So far, I have observed that getting sim2real to work reliably is hard. It requires tweaks that don't scale. Policies that can play perfectly in simulation fall apart because of latencies and the messiness of the real world. These aspects could be modeled to improve the simulation, but not without sinking significant human engineering hours. I have higher hopes for learning from tele-operated data, but that requires a human to learn the task first. These experiments are on my to-do list. I have to learn to play some of the games well through the robot. I’m half-decent at playing Pong and Ms Pacman now. Learning directly on robots is looking like the most promising approach. This approach takes away pesky distribution shifts and makes it possible to have algorithms that continually improve with more data and time without any human intervention. It feels great to let experiments run overnight and wake up to find improved policies. With learning on robots, I should, in principle, be able to go on a long vacation and come back to find better policies for complex tasks beyond Atari games. Whether that is possible with current learning algorithms is a different question.

Khurram Javed

52,110 Aufrufe • vor 8 Monaten