THU-MIG/yoloe

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

2297 stars · 209 forks observed · 2026-10-01

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

YOLOE is the official PyTorch implementation of an open-vocabulary object detection and segmentation model presented at ICCV 2025. It unifies text-prompted, visual-prompted, and prompt-free detection in a single efficient real-time model with zero inference overhead compared to closed-set YOLOs.

Use cases

• detect arbitrary objects in images with text prompts
• zero-shot object detection without retraining
• real-time open-vocabulary instance segmentation
• detect objects using a visual reference prompt
• detect everything in an image without specifying classes
• transfer a detector to new categories with no inference cost

When to choose

• you need real-time open-vocabulary detection and segmentation in one model
• you want YOLO-level speed but with flexible text, visual, or no prompts
• you need zero-shot transfer to unseen object categories
• you want an efficient alternative to heavier open-set detectors like YOLO-World

When to avoid

• you only need fixed closed-set classes and a standard YOLO suffices
• you require a permissive license - it is AGPL-3.0
• you need a non-PyTorch framework or non-GPU deployment
• your use case is classification or captioning rather than detection/segmentation

Facets

library · maturity active

computer-vision image-processing machine-learning deep-learning computer-vision image-processing deep-learning artificial-intelligence python cross-platform object-detection instance-segmentation open-vocabulary zero-shot yolo pytorch iccv-2025 prompt-based-detection gpu

Member repositories

For agents

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