THU-MIG/yolov10

Health v2 · maintenance only

Adoption not part of the score

11336 stars · 1180 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

YOLOv10 is a real-time end-to-end object detection model family that removes NMS post-processing via consistent dual assignments and optimizes architecture for efficiency-accuracy tradeoffs. The repository also introduces YOLOE, an extension supporting open-set detection and segmentation with text, visual, and prompt-free mechanisms.

Use cases

• detect objects in images in real time
• train a custom YOLO object detector on my own dataset
• run NMS-free end-to-end object detection for low-latency deployment
• detect objects from open-vocabulary text prompts
• perform real-time instance segmentation
• deploy an efficient detector on edge devices

When to choose

• you need state-of-the-art real-time object detection with low inference latency
• you want NMS-free end-to-end deployment
• you need open-vocabulary or prompt-free detection and segmentation
• you want pretrained YOLO variants spanning nano to extra-large scales

When to avoid

• your project requires a permissive license (AGPL-3.0 is copyleft)
• you need classification-only or generative vision tasks
• you prefer a mature ecosystem like Ultralytics with broader tooling and integrations

Facets

library · maturity active

computer-vision machine-learning deep-learning image-processing computer-vision machine-learning deep-learning artificial-intelligence python cross-platform object-detection yolo nms-free real-time-inference instance-segmentation open-set-detection neurips-2024 pytorch gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="THU-MIG/yolov10")