YOLO vs YOLOE: Detect ANY Object Without Retraining

Can an object detector find something that isn't in its class list — without retraining? In this video, I compare a standard closed-set YOLO26 model with YOLOE-26, an open-vocabulary detector that lets you change what it looks for at inference time, just by typing text prompts. 💻 Code (free, try it yourself): https://github.com/AbdullahJirjees/yo... 🔍 In this video: • Why normal YOLO can only detect its 80 fixed COCO classes • What YOLOE is and how text prompts replace the fixed class list • "double-decker bus" — a class YOLO doesn't have • YOLO vs YOLOE side by side on the same image • Changing the target live — same image, same model, new prompt • Prompt-free mode: 4,585 built-in categories, no prompts at all • Where it fails: "person holding a phone" • Does open vocabulary replace custom training? (No — here's when to use each) 📚 Resources • Ultralytics YOLOE docs: https://docs.ultralytics.com/models/y... • YOLOE paper (Real-Time Seeing Anything): https://arxiv.org/abs/2503.07465 • What is open-vocabulary detection? (Roboflow): https://roboflow.com/blog/open-vocabu... 🛠️ Tools used: Python, Ultralytics, YOLO26, YOLOE-26, OpenCV 💬 What object would you test with YOLOE? Tell me in the comments — I might try it in the next video. 👍 If this helped, like the video and subscribe for more practical computer vision. #YOLO #YOLOE #ComputerVision #ObjectDetection #AI #MachineLearning #Python #DeepLearning