YOLOE: The Breakthrough AI Model That Helps Machines See Anything in Real Time | TGP#47 скачать в хорошем качестве

YOLOE: The Breakthrough AI Model That Helps Machines See Anything in Real Time | TGP#47

Dive into the future of computer vision with YOLOE: Real-Time Seeing Anything! 🚀 This podcast episode breaks down the groundbreaking model from THU-MIG, set to be unveiled at ICCV 2025. Discover how YOLOE revolutionizes object detection and segmentation by operating efficiently across text prompts, visual inputs, and prompt-free scenarios with zero inference and transferring overhead. Learn about its exceptional zero-shot performance and transferability, outperforming models like YOLO-Worldv2-S on LVIS with 3x less training cost and 1.4x inference speedup, and achieving gains over closed-set YOLOv8-L on COCO with nearly 4x less training time. Whether you're a researcher or an AI enthusiast, understand how YOLOE is pushing the boundaries of adaptable, real-time vision tasks. Check out the official GitHub repository for code and models: https://github.com/THU-MIG/yoloe #YOLOE #ObjectDetection #Segmentation #RealTimeAI #ComputerVision #MachineLearning #OpenSetDetection #DeepLearning #AICV #ICCV2025 #THUMIG