Simpler. Smarter. Further.
United States
Joined February 2014
- Tweets4.7K
- Following61
- Followers10.3K
- Likes7.4K
Pinned Tweet
Ultralytics YOLO Vision 2026 returns September 13 🚀
Join our hybrid global event for the next generation of Ultralytics YOLO, major launches, live demos, breakthrough research, and real-world deployments.
Open vision, built for the real world.
Register now 👉 bit.ly/4fsgH2S
Ultralytics retweeted
How many vehicles are moving through each lane of a busy highway? 🚗
With @ultralytics YOLO26, you can use region-based object counting to detect, track, and count vehicles across multiple predefined zones in real time.
In this demo, each highway region is highlighted with a different color, displaying live vehicle counts as traffic moves through the scene.
#cars #highway #MachineLearning
Drawing every bounding box manually doesn’t scale. 🏷️
With Moondream 3.1 on the Ultralytics Platform, automatically generate object detection annotations for your dataset classes. Review, refine & train with Ultralytics YOLO26.
Read more ➡️ docs.ultralytics.com/platfor…
#Ultralytics #Moondream #DataAnnotation
Ultralytics retweeted
Depth Anything V2 vs. @ultralytics YOLO26 Depth | Speed comparison 🚀
I compared YOLO26m-depth with depth-anything-v2-small on an NVIDIA RTX 3050 using the same input settings. Even though I used the medium YOLO26 model, it delivered nearly 2× the throughput: 22 FPS vs. 11 FPS.
While results will vary depending on hardware and inference settings, this benchmark highlights why YOLO26 depth is a compelling choice for real-time applications where both speed and depth quality matter.
#machinelearningwithpython #DepthEstimation #Research
New tutorial | Bird's-eye view (BEV) with Ultralytics YOLO26 🚗
Learn how to transform camera footage into a top-down view using homography & YOLO26 to detect, track, and visualize vehicle movement.
Watch here ➡️ youtu.be/6_MHChha8Ds
#BirdEye #Homography #MachineLearning
🎃 Spooky season meets computer vision!
This Halloween, we want to know: What's the scariest thing you've encountered while building a computer vision project?
Vote below and reply with your horror stories! 👇
Happy Halloween from Ultralytics! 🎃
50%👻 Bad dataset labels
8%🧟 Model won't converge
8%🕸️ Deployment errors
33%🎃 Endless false positive
12 votes • 3 daysDrawing boxes. Fixing labels. Repeating it 10,000+ times. We replaced that with one click.
Ultralytics Platform uses SAM-powered annotation to generate precise masks and boxes instantly across all five YOLO tasks — already used across 50M images and 150M+ annotations.
Less labeling, more building. From upload to training to production in minutes.
Get started ➡️ platform.ultralytics.com/?ut…
#Ultralytics #Platform #AutoAnnotation #SAM3
One prompt. Every image. Every frame. 🎯
With Ultralytics v8.4.177, SAM now maintains your prompt across entire image folders and videos! Previously, later images or frames reverted to segment-everything. Now, your prompt stays active throughout.
Learn more ➡️ github.com/ultralytics/ultra…
New tutorial | Train Ultralytics YOLO26 on LVIS using Ultralytics Platform 🚀
Learn how to train YOLO26 on a large dataset with 120K images, 1,200 classes & 1.5M+ annotations, and explore how to train, test, export & deploy the trained model directly on the Platform.
Watch here ➡️ youtu.be/VSmdaIWLSM4
#Ultralytics #YOLO26 #LVIS #ComputerVision
African Wildlife is going wild on the Ultralytics Platform. 🦁🌍
The African Wildlife dataset is now among the top 5 most-starred datasets on the Platform!
Explore the data and train your own computer vision models in just a few clicks—all in one place.
Explore the dataset ➡️ platform.ultralytics.com/ult…
#Ultralytics #ComputerVision #AI
Big one for the vision AI community: YOLO26 now runs natively on @AMD hardware via ROCm. 🚀
Train, run, and deploy across Instinct, Radeon, and Ryzen AI. More hardware choice for everyone building real-world computer vision.
If you're in Hong Kong on Nov 29, catch Rick Hou's session at AMD AI DevDay to see it in action.
Ultralytics YOLO now officially supports AMD GPUs. 👀
Developers can train, run, and deploy YOLO26 models across AMD Instinct, Radeon, and Ryzen AI hardware with ROCm software.
Want to see what that looks like in the real world?
Join Rick Hou from @ultralytics at AMD AI DevDay Hong Kong November 29th for Building Real-World AI with Ultralytics YOLO on AMD ROCm.
Read the blog: rocm.blogs.amd.com/ecosystem…
Register for AI DevDay: amd.cventevents.com/event/19…
GPT-6.1 Terra is the new #1 on our VLM Leaderboard for auto-annotation 🏆
0.560 mAP50 across 36 real-world detection datasets, ahead of GPT-6.1 Sol and GPT-6 Astra, at less than half Sol's cost per image.
Try it in Ultralytics Platform ➡️ ultralytics.com/blog/gpt-6-1…
#Ultralytics #GPT61Terra #AutoAnnotation
Extract pixel-level mask data from Ultralytics YOLO26 segmentation results. 🎯
With just a few lines of code, you can access segmentation masks directly from model predictions and use the coordinates or mask tensors in your own computer vision workflows.
This makes it easy to build custom logic for object measurements, shape analysis, region processing, visualization, and downstream analytics.
Learn more ➡️ docs.ultralytics.com/tasks/s…
#Ultralytics #YOLO26 #Segmentation #MachineLearning
Code👇
"""""""
from ultralytics import YOLO
# Load a model
model = YOLO("yolo26n-seg.pt") # load an official model
model = YOLO("path/to/best.pt") # load a custom model
# Predict with the model
results = model("ultralytics.com/images/bus.j…") # predict on an image
# Access the results
for result in results:
xy = result.masks.xy # mask polygons in pixel coordinates
xyn = result.masks.xyn # normalized mask polygons
masks = result.masks.data # binary masks, shape (N,H,W), dtype torch.uint8
"""""""
Analyze a golfer’s swing and body movement with Ultralytics YOLO pose. ⛳
By tracking key body joints throughout each swing, pose estimation can help coaches and players analyze technique, evaluate posture and movement patterns, and track improvements over time using computer vision.
Explore more ➡️ docs.ultralytics.com/tasks/p…
#Ultralytics #PoseEstimation #GolfTraining
⚡ Ultralytics YOLO 🤝 AMD GPUs
At YOLO Vision 2026, we sat down with @AMD to talk ROCm, open-source AI, and making high-performance computer vision more accessible.
From PyTorch to developer resources, discover what's possible. 🚀
🎥 Watch now: youtube.com/watch?v=AKdyYSKn…
Ultralytics retweeted
Workout monitoring using @ultralytics YOLO26 😄
- Leg extension
- Bench press
- Leg press
#workout #MachineLearning #pose
Detect personal protective equipment in real time using Ultralytics YOLO26! 🦺
In this demo, YOLO26 identifies safety gear instantly to support workplace compliance, reduce risk, and enable automated site monitoring in industrial environments.
Read more ➡️ ultralytics.com/blog/using-u…
#Ultralytics #Construction #AI #Research
New tutorial | Speed up Ultralytics YOLO26 inference with Batching 🚀
Learn how to achieve up to a 2× inference speed boost in our benchmark with a batch size of 16 running on @Apple hardware. Explore Python implementation, GPU utilization & performance trade-offs.
Watch here ➡️ youtu.be/YYCfz5OeEow
#Ultralytics #BatchInference #MLOps
Ultralytics retweeted
Zone + count + heatmaps + speed @ultralytics YOLO26 ⚽🔥
💡 Last week, I spent some time adding more graphs for advanced analytics that include line graphs, bar plots, and distribution pie graphs. Some glimpses of the graphs are mentioned in the comments 🙊
More info 👇
#Analytics #MachineLearning #Research
Segment road potholes in real time using Ultralytics YOLO26! 🕳️
Accurately identify potholes in images or video to support road maintenance, safety monitoring, and smart city infrastructure workflows.
Read more ➡️ docs.ultralytics.com/tasks/s…
#AI #SmartCities #Ultralytics #YOLO26