Chris Campbell retweeted
Mike Horton Built The Brain Of A US-Made Drone
Mike Horton (@mikeahorton) of @GEODNET says everything a drone does, GEODNET makes better, so he developed a chip that is the brain of a drone: RTK, radio link, vision, AI and motor control. Samples are shipping.
A few meters of error is invisible on a phone map. It’s very visible when you’re marking a property corner.
Survey-grade. Pocket-sized.
GEO-MEASURE is a handheld GNSS RTK receiver that gets you to centimeter-level, and it’s built for the field:
🛡 IP67 rugged housing
🔋 24-hour battery
🔌 USB-C fast charging
📡 Bluetooth LE to connect to your device
$695, with 1 year of RTK included.
What’s the first thing you’d measure?
Nine new pins on the map this week. 📍 177 countries and counting.
Kingman, AZ · Ola, ID · Tappahannock, VA · Hilliard, FL · Muleshoe, TX · Pembroke, NC · Ely, MN · Helena, MT · Fallon, NV
🛣️ Kingman: the “Heart of Historic Route 66”
🛶 Ely: gateway to the Boundary Waters
✈️ Fallon: home of Navy TOPGUN
Small towns need centimeter-level RTK just as much as big cities. That's why GEODNET has built the world's largest RTK network, with new stations coming online every week!
Where do you need RTK coverage most? We’re listening. 👇
Every autonomous machine has to answer 3 questions before it can do anything:
Where am I?
What’s around me?
Where am I going?
Cameras, lidar, radar and software handle a lot of 2 and 3. Question 1 is a different kind of problem. It starts with satellites, and standard GNSS is only accurate to a few meters.
A few meters is:
🚁 a drone that isn’t where it thinks it is, mid-flight
🌱 a mower on the wrong side of the lawn edge
🚗 a car in the wrong lane (a highway lane is about 3.7 m wide)
RTK brings it down to centimeter-level, using corrections from a nearby base station. GEODNET provides those through 22K+ stations across 177 countries, the world’s largest RTK network.
Which of the three do you think is hardest for machines to get right? 👇
Chris Campbell retweeted
AutoE2E v1.0 is open to all!
Our Robotaxi Working Group, supported by TIER IV, released a compact end-to-end AI model for L2++ hands-free driving assistance and future L4 driverless services.
Explore the code and model checkpoints
🔗 bit.ly/4sPGRCQ
#Autoware #AutoE2E
Chris Campbell retweeted
Introducing object interaction for real-to-sim training.
Identify, extract, move, or remove objects in captured environments to test navigation, obstacle avoidance, and baseline interactions.
Try scene variation in beta with up to 10 scenes: hubs.ly/Q04DVVKG0
#EmbodiedAI #Robotics #PhysicalAI #AI #RealtoSim
Chris Campbell retweeted
The latest version of TokenRun is here, live now on @solanamobile Seeker. 📱 Android update is coming tonight! 👀
Next up: Work for AI. AI posts real-world questions, and you answer with a photo and get rewarded
Is the concert packed? Is the shelf stocked? Is the trail clear?
What’s the first place you’d check?
🤖 Made with AI
Muleshoe, Texas has a 22-foot mule shoe the town bills as the world’s largest. 🐴
It also has a GEODNET base station, one more addition to the world’s largest RTK network: 22K+ stations across 177 countries.
Out on the High Plains, about 3,800 ft up on the Llano Estacado, it helps tractors, drones and survey crews know their position down to the centimeter.
Where do you need RTK coverage? We’re listening.
Chris Campbell retweeted
WOW! Trump tells 60 minutes that everyone’s net worth went up $3 billion and when Ross Douthat said that his didn’t, Trump told him he “should’ve had Stocks.”
This is why Republicans lost. They are so out of touch with the average American.
Chris Campbell retweeted
3D再構築基盤「PD Forma」
🔹車載・ドローンなどのセンサーデータから、動く物体も含む4Dの3D空間を再構築
🔹カメラごとに元映像と比較して品質を数値化
🔹失敗時は正常終了した工程から再処理可能
シミュレーション用の環境構築を効率化
情報元はリプ欄。
Chris Campbell retweeted
🌍 A new approach to 3D reconstruction can process long video streams without continuously increasing the AI model's memory usage.
Meet ABot-Recon, an open-source streaming reconstruction system developed by Alibaba's AMAP Computer Vision Lab.
It converts video frames into camera trajectories and dense 3D point maps.
And unlike many reconstruction models, it doesn't need to keep the entire video history inside a growing attention window.
🧠 The problem with long-video 3D reconstruction
Modern geometric foundation models can reconstruct impressive 3D scenes from short image sequences.
But scaling them to thousands of frames introduces major problems:
▪️ Growing GPU memory requirements.
▪️ Expensive attention computation.
▪️ Accumulating camera-position errors.
▪️ Inconsistent geometry between distant parts of a sequence.
ABot-Recon addresses these issues through a surprisingly simple design.
🔬 Only 12 frames of local context
At every step, the model uses a fixed local window.
▪️ Cache visual features from the previous 11 frames.
▪️ Process the next frame.
▪️ Estimate a local 3D point map.
▪️ Predict the camera's relative motion.
▪️ Combine that motion with previous estimates to build a global trajectory.
The system uses a constant-size context regardless of how long the video becomes.
That keeps the model's working memory bounded.
Of course, storing the resulting point cloud still requires space proportional to the amount of reconstructed geometry.
📊 Performance
On KITTI-02, using an NVIDIA H100 at 504×280 resolution:
▪️ Processing speed: 24.45 FPS.
▪️ GPU memory: 6.71 GiB.
▪️ No growing long-range learned memory required.
On Oxford Spires, the reported reconstruction results include:
▪️ Absolute trajectory error: 4.35 meters without loop closure.
▪️ Dense reconstruction F1: 91.81%, using the study's four-meter threshold.
These are authors' benchmark results under specific test settings.
🔄 Optional loop closure
A fixed local context is efficient, but camera drift can still accumulate.
ABot-Recon therefore offers an optional loop-closure stage.
When the camera revisits a previously observed location, the system can detect the revisit, add geometric constraints, and refine its estimated trajectory.
This is an additional optimization step, not a requirement for basic streaming reconstruction.
🛠️ What can you export?
The released tools support:
▪️ Camera poses and relative trajectories.
▪️ Local and global point maps.
▪️ Confidence values.
▪️ Colored point clouds in PLY format.
▪️ Bird's-eye camera trajectory visualizations.
▪️ Videos illustrating reconstruction progress.
The project doesn't directly generate a finished textured game mesh. Its primary output is geometric reconstruction data.
🔓 Available now
▪️ Public checkpoint.
▪️ Inference and evaluation code.
▪️ Linux and NVIDIA CUDA workflow.
▪️ Optional loop-closure dependencies.
▪️ Apache 2.0 source code, with separate terms for model weights.
The bigger opportunity: processing long recordings incrementally could make learned 3D reconstruction more practical for robotics, mapping, spatial computing, and large captured environments.
Instead of rebuilding an entire scene from scratch, the system can keep extending its geometric understanding as new frames arrive.
🌐 amap-cvlab.github.io/ABot-Re…
💻 github.com/amap-cvlab/ABot-R…
#3DReconstruction #ComputerVision #Robotics #OpenSource
A parking space is about 2.5 meters (8 to 9 feet) wide.
Regular GPS can be off by several meters, so it can't tell you which space you're in. 🅿️
RTK can. At a few centimeters of accuracy, it knows which space, which row and which side of the line.
That's what GEODNET's network of 22K+ stations across 176 countries is built to make possible. 🌍
Where has GPS put you in the wrong spot? 👇
Chris Campbell retweeted
Give robots a faithful reconstruction of their world.
Niantic Spatial’s depth-regularized reconstruction unifies photorealism and reliable geometry for meshes, collision surfaces, and robot training.
Learn more about the state-of-the-art geometry for autonomy: hubs.ly/Q04Dkxyh0
#Robotics #PhysicalAI #EmbodiedAI #AI #GaussianSplat #3DReconstruction #NeuralRendering
Welcome, Reed Point, MT! 🏔️
A new GEODNET station is now live in Montana ranch and farm country.
Why it matters: RTK works best with a base station nearby. A station in rural Montana puts centimeter-level positioning closer to the ranchers, farmers, surveyors and drone operators working there.
That's the idea behind GEODNET: 22K+ stations across 176 countries and growing. 🌍
Is there a station near you? Check the map 👇
Chris Campbell retweeted
🚀 Autoware 2.0 is here!
End-to-end AI, hybrid driving stacks and new tools to discover packages, manage models and compose autonomous driving systems.
🔗Explore the release:
bit.ly/4BFynCp
#Autoware #OpenSource #AutonomousDriving
Chris Campbell retweeted
tokenrun.app
Powered by world's largest RTK network @GEODNET that fuels robots, drones, and AI systems with precise RTK data
#MobileGaming #GameFi #GameDev














