Y

Yoloe

Developed by jameslahm
YOLOE is an efficient, unified, and open model for object detection and segmentation, supporting various prompting mechanisms including text, visual inputs, and prompt-free paradigms, achieving real-time universal visual perception.
Downloads 40.34k
Release Time : 3/10/2025

Model Overview

YOLOE integrates detection and segmentation functions under multiple open prompting mechanisms into a single efficient model. For text prompts, it proposes a reparameterizable region-text alignment strategy; for visual prompts, it designs a semantically activated visual prompt encoder; and for prompt-free scenarios, it develops a lazy region-prompt contrast strategy.

Model Features

Multiple Prompting Mechanisms
Supports various prompting mechanisms including text prompts, visual inputs, and prompt-free paradigms
Efficient Real-time Processing
Achieves real-time visual perception while maintaining high inference efficiency and low training costs
Reparameterizable Design
Proposes a reparameterizable region-text alignment strategy to achieve zero inference and transfer overhead
Open Scene Adaptation
Breaks through the predefined category limitations of traditional YOLO models to adapt to open scenes

Model Capabilities

Object detection
Image segmentation
Text prompt recognition
Visual prompt recognition
Prompt-free object recognition

Use Cases

Intelligent Surveillance
Open Scene Object Recognition
Recognizes various objects in surveillance scenes without predefined category limitations
Accurately identifies various objects, including rare or newly appeared ones
Autonomous Driving
Real-time Road Object Detection
Detects various objects on the road in real-time for autonomous driving systems
High-precision identification of various traffic participants with fast processing speed
Industrial Quality Inspection
Defect Detection
Identifies product defects through visual prompts
Adapts to defect detection needs for different types of products
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase