Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

This is next level for hardware! Robots don't always have room for a camera built for someone else's machine. Luxonis | Robotic Vision just launched OAK 4 FFC, the first OAK Modular board on OAK 4. You pick up to five cameras and put them wherever the robot needs...

Scobleizer's profile picture

Robert Scoble

616,629 subscribers

10,500 Aufrufe • vor 7 Tagen •via X (Twitter)

Wissenschaft & TechnologieAutos & Fahrzeuge
LIVE

Anya Rossi• Live Now

Private livecam show

16 Kommentare

Profilbild von Josh Marino 🦞🦾🥩
Josh Marino 🦞🦾🥩vor 7 Tagen

@luxonis Tesla built FSD on this same idea: several cameras captured in sync, fused into one bird's-eye view. Now a small humanoid or drone team gets five synced cameras, stereo depth and on-board stitching for $849. Camera-first robot vision just got a lot cheaper to try.

Profilbild von James Malsawm
James Malsawmvor 7 Tagen

@luxonis this is impressive

Profilbild von Tinker
Tinkervor 7 Tagen

@luxonis Up to five cameras on one board matters most for robot arms and small rovers, where a camera has to sit on the gripper or low on the chassis. I'd like to know how they handle sync and calibration across all five.

Profilbild von Crypto Cobra
Crypto Cobravor 7 Tagen

@luxonis Robert knows ball👀🖤

Profilbild von ᴋᴀᴋᴀʀᴏᴛ
ᴋᴀᴋᴀʀᴏᴛvor 7 Tagen

@luxonis Synchronized multi camera vision with onboard AI could be useful for robots operating in complex environments.

Profilbild von Yueke
Yuekevor 7 Tagen

@luxonis Modular makes total sense here — every robot really does need to see differently.

Profilbild von Anish
Anishvor 7 Tagen

@luxonis this is such a good video... the storytelling and production is spot on

Profilbild von Japan Gor
Japan Gorvor 7 Tagen

@luxonis The sync-on-demand is actually useful for robot frames.

Profilbild von MORI⚡️
MORI⚡️vor 7 Tagen

@luxonis that modular approach from luxonis makes so much sense for robots with tight spaces

Profilbild von UncleRez
UncleRezvor 7 Tagen

@luxonis The synchronized cameras are the part that really got me Getting multiple angles is one thing but making them work together in real time is a whole different beast Wonder how well it handles fast-moving scenes

Profilbild von Leo | Building KOKO
Leo | Building KOKOvor 7 Tagen

@luxonis Where the cameras go matters more than people think. In the aging parents' homes where we test KOKO, a lot of what it needs to see is low: reading glasses left on a side table, a dropped phone by the armchair. Curious where teams are mounting OAK 4 on home robots.

Profilbild von The AI Grandpa@theaigrandpa
The AI Grandpa@theaigrandpavor 7 Tagen

@luxonis Five cameras on one tiny board and the robot still sees more than I do before coffee.

Profilbild von Gill
Gillvor 7 Tagen

@luxonis Dynamic calibration without physical targets solves a massive headache

Profilbild von Zuri 🇵🇹
Zuri 🇵🇹vor 7 Tagen

@luxonis tiny board with big brain

Profilbild von Brent
Brentvor 6 Tagen

@luxonis WHOA !!!! 😳

Profilbild von Michael Mizrahi
Michael Mizrahivor 7 Tagen

@luxonis Could Grok or FSD leverage something like this for car-to-car intelligent routing as part of Tesla’s homegrown maps?

Ähnliche Videos

🚨 BREAKING: Big news in the computer vision world! 🎥 Luxonis | Robotic Vision just dropped its new OAK 4 line, and it’s a big upgrade for edge computer vision. Instead of being “just a stereo camera,” OAK 4 is a fully standalone vision computer with 52 TOPS of on-device AI. Models run locally, depth is computed locally, and no external PC or cloud pipeline is required. This is why robotics teams love it: lower latency, lower cost, fewer failure points in the field. The hardware is built for the real-world. IP67, shock-resistant, wide-FOV RGB + stereo pair, IR projection, IMU, audio, and a patent-pending calibration system that keeps depth accurate even when conditions change. But the real move is the platform. With Luxonis Hub, you can deploy models, grab telemetry, push OTA updates, or collect data when performance drifts, all from a unified interface. It turns a single device into an end-to-end edge CV system. Most customers today in robotics are groups who just want something that works: AMRs, bin-picking systems, trailer-loading robots, and ag-tech. 🤖 And they all say the same thing, the appeal isn’t raw TOPS, it’s the all-in-one simplicity that lets them scale without building custom infrastructure. Feels like the direction edge vision has been waiting for: rugged hardware + high-throughput on-device compute + a real management layer. A next step toward “plug-and-deploy” perception for robots. 🔗 Find out more here: ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →
1:12

🚨 BREAKING: Big news in the computer vision world! 🎥 Luxonis | Robotic Vision just dropped its new OAK 4 line, and it’s a big upgrade for edge computer vision. Instead of being “just a stereo camera,” OAK 4 is a fully standalone vision computer with 52 TOPS of on-device AI. Models run locally, depth is computed locally, and no external PC or cloud pipeline is required. This is why robotics teams love it: lower latency, lower cost, fewer failure points in the field. The hardware is built for the real-world. IP67, shock-resistant, wide-FOV RGB + stereo pair, IR projection, IMU, audio, and a patent-pending calibration system that keeps depth accurate even when conditions change. But the real move is the platform. With Luxonis Hub, you can deploy models, grab telemetry, push OTA updates, or collect data when performance drifts, all from a unified interface. It turns a single device into an end-to-end edge CV system. Most customers today in robotics are groups who just want something that works: AMRs, bin-picking systems, trailer-loading robots, and ag-tech. 🤖 And they all say the same thing, the appeal isn’t raw TOPS, it’s the all-in-one simplicity that lets them scale without building custom infrastructure. Feels like the direction edge vision has been waiting for: rugged hardware + high-throughput on-device compute + a real management layer. A next step toward “plug-and-deploy” perception for robots. 🔗 Find out more here: ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Lukas Ziegler

41,955 Aufrufe • vor 10 Monaten

OAK 4 CS now supports PTP synchronization. Sync multi-camera exposures over standard Ethernet. No FSYNC wiring, no daisy chains, no trigger cables across long distances. Just a shared clock distributed over the network. For multi-camera perception, distributed AI vision, and sensor fusion at scale.
0:43

OAK 4 CS now supports PTP synchronization. Sync multi-camera exposures over standard Ethernet. No FSYNC wiring, no daisy chains, no trigger cables across long distances. Just a shared clock distributed over the network. For multi-camera perception, distributed AI vision, and sensor fusion at scale.

Luxonis | Robotic Vision

32,787 Aufrufe • vor 5 Monaten

Today we're announcing microSLAM, a monocular SLAM system built by our computer vision team in Zürich with ETH Zurich's Computer Vision and Geometry Lab. It ranks first on the LaMaria benchmark for monocular SLAM, and holds up against systems that carry many more cameras and IMUs. Ours runs on a single RGB stream. Robot learning is bottlenecked on data, and the largest untapped source is people going about their work. A camera on someone's head for an afternoon in a plant is a record of how that plant actually operates, including all the parts nobody writes down. That data is only useful if you can recover the geometry, and geometry from one moving camera in a world that will not sit still is the hard version of the problem. Solving it monocular is what makes the data cheap. Rigs do not scale to every worker. Glasses do. microSLAM turns that footage into interactive 3D environments where behavioral foundation models can be trained. One person walks a factory floor. A robot learns to navigate it.
3:30

Today we're announcing microSLAM, a monocular SLAM system built by our computer vision team in Zürich with ETH Zurich's Computer Vision and Geometry Lab. It ranks first on the LaMaria benchmark for monocular SLAM, and holds up against systems that carry many more cameras and IMUs. Ours runs on a single RGB stream. Robot learning is bottlenecked on data, and the largest untapped source is people going about their work. A camera on someone's head for an afternoon in a plant is a record of how that plant actually operates, including all the parts nobody writes down. That data is only useful if you can recover the geometry, and geometry from one moving camera in a world that will not sit still is the hard version of the problem. Solving it monocular is what makes the data cheap. Rigs do not scale to every worker. Glasses do. microSLAM turns that footage into interactive 3D environments where behavioral foundation models can be trained. One person walks a factory floor. A robot learns to navigate it.

Gianni Hodel

74,497 Aufrufe • vor 1 Monat

Football is so great. Oak Ridge backup QB Kipton Feagan (Kipton Feagan) had to enter the game late. In OT...4th down...game on the line...and found Zayden Williams (4⭐️ CB Zayden Williams 2028) in the endzone to win it. Huge win for Oak Ridge Football to open the season!
0:39

Football is so great. Oak Ridge backup QB Kipton Feagan (Kipton Feagan) had to enter the game late. In OT...4th down...game on the line...and found Zayden Williams (4⭐️ CB Zayden Williams 2028) in the endzone to win it. Huge win for Oak Ridge Football to open the season!

Sam Rothman

15,329 Aufrufe • vor 2 Monaten

I can't f*ckin believe it - I BUILT a cat-inspired visual system for a ROBOT DOG A camera feeds visual data into a simulated feline neural system - and its activity controls where the robot moves. → 20,480 simulated neurons processing vision → 8 orientation channels detecting edges → Camera → NEURAL ACTIVITY → target → movement → Turns toward moving targets in REAL TIME → Runs completely OFFLINE → Connected to a Unitree Go2 through its SDK Movement appears right - it turns right. Left - it turns left. Centered - it locks on. The robot’s controller handles balance and legs. CATBRAIN handles where it looks and follows. The project is still IN DEVELOPMENT - you can follow every step on GitHub. I attached the link in the replies.
0:39

I can't f*ckin believe it - I BUILT a cat-inspired visual system for a ROBOT DOG A camera feeds visual data into a simulated feline neural system - and its activity controls where the robot moves. → 20,480 simulated neurons processing vision → 8 orientation channels detecting edges → Camera → NEURAL ACTIVITY → target → movement → Turns toward moving targets in REAL TIME → Runs completely OFFLINE → Connected to a Unitree Go2 through its SDK Movement appears right - it turns right. Left - it turns left. Centered - it locks on. The robot’s controller handles balance and legs. CATBRAIN handles where it looks and follows. The project is still IN DEVELOPMENT - you can follow every step on GitHub. I attached the link in the replies.

BuBBliK

44,188 Aufrufe • vor 1 Monat

Robotics and CV! 👓 Computer vision gives the robots a sense of vision and additional intelligence. Pralines are packed by a delta robot using computer vision. Packing chocolate pralines into a box with the delta robot. Unlike most robots, this one has a fixed TCP, and the base moves in response to vision data. To adjust the base, the vision system determines exactly where the praline is located on the conveyor belt. 🍫 Using this application, the system knows where each praline should be placed based on its orientation. A true engineering showcase!
0:36

Robotics and CV! 👓 Computer vision gives the robots a sense of vision and additional intelligence. Pralines are packed by a delta robot using computer vision. Packing chocolate pralines into a box with the delta robot. Unlike most robots, this one has a fixed TCP, and the base moves in response to vision data. To adjust the base, the vision system determines exactly where the praline is located on the conveyor belt. 🍫 Using this application, the system knows where each praline should be placed based on its orientation. A true engineering showcase!

Lukas Ziegler

79,438 Aufrufe • vor 1 Jahr

🤯 A hidden gem in the latest vivo X300 Ultra update! The new Telephoto Calibration feature is more than just an animation. Turn it on in Camera Settings and it asks you to flip the phone over. At first, I thought nothing was happening… until I looked closely at the periscope lens. The telephoto camera was actually moving and adjusting itself in real time for calibration! A phone that can visually calibrate its own periscope lens for optimal performance? That's a proper wow moment 📸🔥 vivo is doing some really interesting things in the camera department.
0:44

🤯 A hidden gem in the latest vivo X300 Ultra update! The new Telephoto Calibration feature is more than just an animation. Turn it on in Camera Settings and it asks you to flip the phone over. At first, I thought nothing was happening… until I looked closely at the periscope lens. The telephoto camera was actually moving and adjusting itself in real time for calibration! A phone that can visually calibrate its own periscope lens for optimal performance? That's a proper wow moment 📸🔥 vivo is doing some really interesting things in the camera department.

S M A Sithick

25,336 Aufrufe • vor 4 Monaten

Erik Voorhees on the crypto UX realisation that took 15 years "Crypto has always had this UX challenge with humans. Every time you go to a crypto conference, whether it was back in 2011 or tomorrow, people talk about how do you make this stuff more usable to people" "It seems like perhaps it's just much more usable for robots and the machines and the AI, and that crypto is actually built for the machines and we didn't know it. The machines are now going to start running on the stuff and utilizing it" "They have no trouble with public key pairs, that's easy for them" "I don't see a future where the robots are using bank accounts at Wells Fargo. I do see a future where they're using crypto assets on decentralized rails"
0:48

Erik Voorhees on the crypto UX realisation that took 15 years "Crypto has always had this UX challenge with humans. Every time you go to a crypto conference, whether it was back in 2011 or tomorrow, people talk about how do you make this stuff more usable to people" "It seems like perhaps it's just much more usable for robots and the machines and the AI, and that crypto is actually built for the machines and we didn't know it. The machines are now going to start running on the stuff and utilizing it" "They have no trouble with public key pairs, that's easy for them" "I don't see a future where the robots are using bank accounts at Wells Fargo. I do see a future where they're using crypto assets on decentralized rails"

Laura Shin

49,927 Aufrufe • vor 2 Monaten

stereo depth map on the orange pi is now at 18-22.4 FPS all of the compute is happening on the orange pi zero, including camera capture. it is the board on the bottom with a quarter for scale.
0:59

stereo depth map on the orange pi is now at 18-22.4 FPS all of the compute is happening on the orange pi zero, including camera capture. it is the board on the bottom with a quarter for scale.

kache

17,792 Aufrufe • vor 2 Monaten

Ex-OpenAI safety researcher Steven Adler says AI labs are monitoring their models like a bank that lets the robber turn off the security cameras: "Imagine you ran a bank, and you want the bank to not get robbed. The way you do this is not setting up a security camera and every hour you check the feed to see if the bank was robbed in that time, and then you try to respond." "You certainly don't leave it where a criminal could walk into the bank and turn off the camera, and that's the equivalent of what's happening at AI labs today, not just in terms of automated AI R&D and recursive self-improvement, but broadly across the board."
1:45

Ex-OpenAI safety researcher Steven Adler says AI labs are monitoring their models like a bank that lets the robber turn off the security cameras: "Imagine you ran a bank, and you want the bank to not get robbed. The way you do this is not setting up a security camera and every hour you check the feed to see if the bank was robbed in that time, and then you try to respond." "You certainly don't leave it where a criminal could walk into the bank and turn off the camera, and that's the equivalent of what's happening at AI labs today, not just in terms of automated AI R&D and recursive self-improvement, but broadly across the board."

MTS

16,470 Aufrufe • vor 2 Monaten

Angela White admits that sex on camera has no intimacy. She says she just has to 'open up for the light' so the camera can 'see the penetration' and follow a 'script' with a 'performative aspect. "When you're having sex for work, you have to open up for the light, you have to open up so the camera can see the penetration... there's a script, there's a performative aspect."
0:46

Sensitive content

Angela White admits that sex on camera has no intimacy. She says she just has to 'open up for the light' so the camera can 'see the penetration' and follow a 'script' with a 'performative aspect. "When you're having sex for work, you have to open up for the light, you have to open up so the camera can see the penetration... there's a script, there's a performative aspect."

Rixlerr

1,447,116 Aufrufe • vor 2 Monaten

🚨 THE CEO OF FLOCK JUST SAID THE QUIET PART OUT LOUD — AND PEOPLE ARE LOSING IT Remember when Flock cameras were only supposed to read license plates? Apparently that's not enough anymore. Now we're talking about AI that can find a vehicle from a simple description. Microphones that are always listening for so called "sounds of distress." Drones that can be launched automatically after a 911 call. And camera networks that can follow movement across entire cities. Every year the cameras get smarter. The databases get bigger. The AI gets more powerful. And the amount of information being collected keeps growing. Then came the comment that really got people's attention. The CEO behind Flock, one of the largest camera networks in America reportedly compared people who map camera locations to terrorists. Think about that for a second. The cameras aren't the problem. The people tracking the cameras are. Some people see a tool that makes communities safer. Others see a system that knows where you've been, where you are, and eventually where you're likely to go next. If mapping Flock cameras makes you a "terrorist," what does that make the people putting cameras everywhere?
6:18

🚨 THE CEO OF FLOCK JUST SAID THE QUIET PART OUT LOUD — AND PEOPLE ARE LOSING IT Remember when Flock cameras were only supposed to read license plates? Apparently that's not enough anymore. Now we're talking about AI that can find a vehicle from a simple description. Microphones that are always listening for so called "sounds of distress." Drones that can be launched automatically after a 911 call. And camera networks that can follow movement across entire cities. Every year the cameras get smarter. The databases get bigger. The AI gets more powerful. And the amount of information being collected keeps growing. Then came the comment that really got people's attention. The CEO behind Flock, one of the largest camera networks in America reportedly compared people who map camera locations to terrorists. Think about that for a second. The cameras aren't the problem. The people tracking the cameras are. Some people see a tool that makes communities safer. Others see a system that knows where you've been, where you are, and eventually where you're likely to go next. If mapping Flock cameras makes you a "terrorist," what does that make the people putting cameras everywhere?

HustleBitch

320,152 Aufrufe • vor 4 Monaten

Drop the camera, and bang -- the robots do the job without camera calibration or additional data collection. We want robots to manipulate anywhere! Our new work, **Maniwhere**, enables robots to manipulate objects from any camera view with any background. Isn't this the generalization every roboticist is looking for? How do we achieve this with sim2real? You can check this thread from Zhecheng Yuan ! And the website is here:
0:29

Drop the camera, and bang -- the robots do the job without camera calibration or additional data collection. We want robots to manipulate anywhere! Our new work, **Maniwhere**, enables robots to manipulate objects from any camera view with any background. Isn't this the generalization every roboticist is looking for? How do we achieve this with sim2real? You can check this thread from Zhecheng Yuan ! And the website is here:

Huazhe Harry Xu

28,935 Aufrufe • vor 2 Jahren

I became an Acorns investor, because I believe in financial freedom and making money moves as smart as possible. That’s why Acorns and I built the Mighty Oak Debit Card for you. Acorns don’t grow into mighty oaks overnight. They grow little by little, and bit by bit. Same is true with growing your money — it doesn’t happen overnight. Save a little, invest a little, and over time, your oak could grow too. That’s what Acorns and the Mighty Oak Debit Card enable. The Mighty Oak Debit Card gives you a way to save and invest every time you make a purchase, get your paycheck, shop online, and more. I’m right there with you and I know that every cent matters and every paycheck is hard earned. Head over to or download the Acorns app to sign up for your Mighty Oak Debit Card, and start saving and investing NOW. I wish you great growth! Slow and mighty- #GrowYourOak #Acorns #Investor
1:13

I became an Acorns investor, because I believe in financial freedom and making money moves as smart as possible. That’s why Acorns and I built the Mighty Oak Debit Card for you. Acorns don’t grow into mighty oaks overnight. They grow little by little, and bit by bit. Same is true with growing your money — it doesn’t happen overnight. Save a little, invest a little, and over time, your oak could grow too. That’s what Acorns and the Mighty Oak Debit Card enable. The Mighty Oak Debit Card gives you a way to save and invest every time you make a purchase, get your paycheck, shop online, and more. I’m right there with you and I know that every cent matters and every paycheck is hard earned. Head over to or download the Acorns app to sign up for your Mighty Oak Debit Card, and start saving and investing NOW. I wish you great growth! Slow and mighty- #GrowYourOak #Acorns #Investor

Dwayne Johnson

602,651 Aufrufe • vor 3 Jahren

69% of baseball fans say they'd rather have a computer vision AI system call balls and strikes than a human umpire. This season, the MLB gave them one. For the first time in league history, a human ump's ball-strike call is not final. A Sony computer vision system called Hawk-Eye makes the ruling. It can read the seam pattern on the ball, measure spin axis, and detect spin decay mid-flight. Hawk-Eye's Head of Computer Vision Engineering says the pipeline runs "various AI and machine learning models" from camera capture to data output. On Saturday, one umpire had 6 of 8 challenged calls overturned. Three of those missed by over 2 inches. The team ran out of challenges by the fourth inning trying to correct him, then the manager got thrown out for arguing a call they couldn't challenge. An AI computer vision system accurate to a sixth of an inch, sitting next to a human who misses by two. The crowd cheered for the machine.
1:26

69% of baseball fans say they'd rather have a computer vision AI system call balls and strikes than a human umpire. This season, the MLB gave them one. For the first time in league history, a human ump's ball-strike call is not final. A Sony computer vision system called Hawk-Eye makes the ruling. It can read the seam pattern on the ball, measure spin axis, and detect spin decay mid-flight. Hawk-Eye's Head of Computer Vision Engineering says the pipeline runs "various AI and machine learning models" from camera capture to data output. On Saturday, one umpire had 6 of 8 challenged calls overturned. Three of those missed by over 2 inches. The team ran out of challenges by the fourth inning trying to correct him, then the manager got thrown out for arguing a call they couldn't challenge. An AI computer vision system accurate to a sixth of an inch, sitting next to a human who misses by two. The crowd cheered for the machine.

The Rundown AI

198,613 Aufrufe • vor 7 Monaten

A generalist model, not a dedicated robot policy, pulled off physical control by reasoning and using tools. Here's the four-step approach Gemini 4 figured out, given high-level instructions and the overall task: 1. Watched a human throw, and mapped the motion to its own joints. 2. Used its cameras to locate the board in 3D, and built a physics sim. 3. Practiced in simulation, tuning its stance and swing for a smooth arc. 4. On the real robot, watched each landing and adjusted by centimeters until 7 shots went in.
0:37

A generalist model, not a dedicated robot policy, pulled off physical control by reasoning and using tools. Here's the four-step approach Gemini 4 figured out, given high-level instructions and the overall task: 1. Watched a human throw, and mapped the motion to its own joints. 2. Used its cameras to locate the board in 3D, and built a physics sim. 3. Practiced in simulation, tuning its stance and swing for a smooth arc. 4. On the real robot, watched each landing and adjusted by centimeters until 7 shots went in.

The Humanoid Hub

10,160 Aufrufe • vor 21 Tagen

STEAM FRAME UPDATE The Arcturus Vision Camera for Steam Frame now has an “Auto night vision” toggle, where the software automatically detects when lights are turned off and switches to your built in camera for night vision.
0:42

STEAM FRAME UPDATE The Arcturus Vision Camera for Steam Frame now has an “Auto night vision” toggle, where the software automatically detects when lights are turned off and switches to your built in camera for night vision.

Steam Hardware Updates

23,316 Aufrufe • vor 24 Tagen

The robot’s “eyes” just received a big upgrade. LingBot-Depth 2.0, a depth-completion model with half the depth error just dropped. 12/16 benchmarks topped. Glass, mirrors, and transparent objects are so easy for us humans, but so hard for robots, because they do not behave like ordinary surfaces in a camera pipeline. A robot that misunderstands a balcony window or a table edge, will have a completely false planning inside a false world. Huge implecation. LingBot-Depth 2.0 takes an RGB image plus a broken depth map from a sensor and then outputs a cleaner depth map and a usable 3D point cloud. Numbers on LingBot-Depth 2.0 - Excels on glass, mirrors & transparent objects — where traditional depth cameras fail - Training data: 3M → 150M (50x scale-up) - 12 out of 16 first-place rankings on depth completion benchmarks - RMSE cut in half: 0.132 → 0.062 on the hardest indoor scenes LingBot-Vision trained on boundaries, because object edges carry the geometry robots need. No human boundary labels are used, which makes this approach easier to scale. The open-sourced LingBot-Vision is the general vision backbone, and LingBot-Depth 2.0 is the depth model built on it.
2:22

The robot’s “eyes” just received a big upgrade. LingBot-Depth 2.0, a depth-completion model with half the depth error just dropped. 12/16 benchmarks topped. Glass, mirrors, and transparent objects are so easy for us humans, but so hard for robots, because they do not behave like ordinary surfaces in a camera pipeline. A robot that misunderstands a balcony window or a table edge, will have a completely false planning inside a false world. Huge implecation. LingBot-Depth 2.0 takes an RGB image plus a broken depth map from a sensor and then outputs a cleaner depth map and a usable 3D point cloud. Numbers on LingBot-Depth 2.0 - Excels on glass, mirrors & transparent objects — where traditional depth cameras fail - Training data: 3M → 150M (50x scale-up) - 12 out of 16 first-place rankings on depth completion benchmarks - RMSE cut in half: 0.132 → 0.062 on the hardest indoor scenes LingBot-Vision trained on boundaries, because object edges carry the geometry robots need. No human boundary labels are used, which makes this approach easier to scale. The open-sourced LingBot-Vision is the general vision backbone, and LingBot-Depth 2.0 is the depth model built on it.

Rohan Paul

25,312 Aufrufe • vor 4 Monaten