S

Sam Vit Large

Developed by Xenova
Large-scale image segmentation model based on Vision Transformer architecture, capable of generating high-quality object masks from input points
Downloads 34
Release Time : 5/31/2023

Model Overview

Segment Anything Model (SAM) is a versatile image segmentation model that automatically generates precise object masks based on user-provided input points (e.g., clicks). Built on Vision Transformer architecture, it exhibits strong zero-shot transfer capabilities.

Model Features

Zero-shot Segmentation Capability
Handles various image segmentation tasks without domain-specific training
Interactive Segmentation
Generates precise object masks guided by simple input points
High-quality Output
Produces refined object boundaries and multiple candidate masks
Web Compatibility
Provides ONNX-format weights for browser environment execution

Model Capabilities

Interactive Image Segmentation
Object Mask Generation
Multi-candidate Output
Zero-shot Image Understanding

Use Cases

Image Editing
Object Removal & Replacement
Generates precise masks via simple clicks for photo editing
Achieves accurate object isolation effects
Computer Vision Annotation
Semi-automatic Data Labeling
Significantly reduces manual annotation workload
3-5x improvement in labeling efficiency
AR/VR Applications
Real-time Object Segmentation
Separates foreground objects in augmented reality scenarios
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase