Ultralytics

The Retail Stall Segmentation dataset is live on Ultralytics Platform, and it's a fascinating one! 🛍️ 1,200 images. 2,867 annotations. Full instance segmentation support. This is the kind of dataset that powers real retail intelligence, space utilization, vendor placement & foot traffic analysis. Though not a proof-of-concept, it is commercially-ready data for the backbone of modern retail operations. Thank you juan-uriel-legaria-pena, for building this and sharing it openly with the community. Contributions like this raise the bar for everyone. 🙏 Ready to put it to work? 👉 Explore the Retail Stall Segmentation dataset on Ultralytics Platform: https://bit.ly/4v6MyNh #ComputerVision #RetailAI #InstanceSegmentation #Ultralytics #OpenData #SmartRetail

Monitor parking lots in real time with Ultralytics YOLO 🚗 In this demo, YOLO26 detects vehicles and monitors parking occupancy across multiple spaces in real time. The system automatically identifies which spots are occupied and which are available, giving operators a live overview of parking activity. This helps reduce traffic congestion, direct drivers to available spots more efficiently, and improve parking management in busy environments such as malls, office buildings, and public parking areas. By turning video streams into actionable insights, computer vision enables smarter, faster, and more efficient parking operations. Learn more ➡️ https://bit.ly/3Nb2s8D #Ultralytics #ParkingManagement #YOLO26

New release v8.4.65 | Mobile inference performance boost 📱⚡ Ultralytics YOLO `v8.4.65` makes mobile deployment faster and smoother with improved QNN exports for Snapdragon NPUs and quicker semantic segmentation outputs on QNN and CoreML. Minor updates: ✅ RKNN export improvements with FP16 support and better INT8 test coverage ✅ More reliable Docker builds and cleaner CI workflows ✅ Clearer export, tracking, and deployment docs Ultralytics v8.4.65 release notes ➡️ https://lnkd.in/eUDTbcpB

New tutorial | LightOn-OCR: VLM-based optical character recognition 📄 Optical character recognition is evolving beyond traditional text extraction pipelines. In this tutorial, we explore LightOn-OCR, designed to balance speed and accuracy when extracting text from images. What we'll cover: ✅ Introduction to LightOn-OCR ✅ Performance and speed comparison with other OCR approaches ✅ OCR test cases across different document layouts ✅ Python implementation and inference workflow ✅ Evaluating extracted text accuracy on real examples A practical walkthrough for developers exploring modern OCR solutions for document processing, automation, and Vision AI workflows. Watch now ➡️ https://lnkd.in/euigGHYB #Ultralytics #OCR #ComputerVision

Track segmented objects in real time using FastSAM. ⚡ FastSAM isn’t just about segmentation; you can also use object tracking alongside it for tracking the segmented objects in real time. By combining fast segmentation with tracking, you can follow objects across frames while maintaining precise outlines. This is especially useful for applications such as video analytics, surveillance, and interactive vision systems, where both speed and accuracy matter. Learn more ➡️ https://bit.ly/4rpV4oG #Ultralytics #FastSAM #ObjectTracking

New release v8.4.63 | TensorRT 11 export support is here 🚀 Ultralytics YOLO `v8.4.63` restores TensorRT 11 export with FP16 and INT8 quantization, adds more built-in tracking options, and improves deployment reliability on modern NVIDIA systems. Minor updates: ✅ Added OC-SORT, Deep OC-SORT, FastTracker, and TrackTrack ✅ Improved stream cleanup for safer multi-stream apps ✅ Added Rust inference docs and refreshed tracking guidance Ultralytics v8.4.63 release notes ➡️ https://lnkd.in/e9ntb5NX

Ultralytics Live Session: Vision AI for aerospace 🚀 Join us on June 23rd at 16:00 CEST to learn how aerospace teams are using Ultralytics YOLO to improve inspection, quality control, and maintenance operations. We’ll cover: ✅ Real-world use cases including aircraft defect detection, manufacturing quality control, and MRO workflows ✅ Best practices for deploying Vision AI in aerospace environments ✅ Live Q&A Register now ➡️ https://bit.ly/4v1EEEV

Monitor power line towers with Ultralytics YOLO26. ⚡ In this demo, YOLO26 is used to detect key components of electric transmission towers, including insulator strings, crossarms, and turret structures, directly from visual data. Instead of manual inspections, computer vision enables continuous monitoring of critical infrastructure, helping teams identify missing or damaged parts early. This is especially useful for large-scale energy networks where regular inspection is challenging. Learn more ➡️ https://bit.ly/4wa8how #Ultralytics #YOLO26 #Energy

New release v8.4.62 | Reliable training checkpoint saves 🛡️ Ultralytics YOLO `v8.4.62` prevents trained models from being lost at the end of training, especially in AMP runs, while improving docs accuracy and overall test reliability. Minor updates: ✅ Safer AMP validation without poisoning EMA weights ✅ Refreshed Ultralytics Platform docs https://lnkd.in/eDMYAFFM and expanded API coverage ✅ More stable tests, leaner CI, and fixed COCO evaluation links Ultralytics v8.4.62 release notes ➡️ https://lnkd.in/e6quGZUv