R

Robustsam Vit Base

Developed by jadechoghari
RobustSAM is a model for robust segmentation on degraded images, improved upon SAM to enhance segmentation performance on low-quality images.
Downloads 314
Release Time : 8/16/2024

Model Overview

RobustSAM is an improved version of the Segment Anything Model (SAM), focusing on maintaining segmentation performance under image quality degradation. It retains SAM's promptability and zero-shot generalization capabilities while optimizing performance on low-quality images with minimal parameter additions and computational requirements.

Model Features

Degraded image robustness
Specifically optimized for segmentation performance on low-quality images (e.g., blur, haze, low light, etc.)
Efficient optimization
Requires only minimal parameter additions and can be optimized within 30 hours on 8 GPUs
Zero-shot capability
Maintains SAM's powerful zero-shot segmentation ability without task-specific training
Prompt system
Supports various prompt methods like points and bounding boxes for flexible segmentation control

Model Capabilities

Image segmentation
Zero-shot segmentation
Prompt-based segmentation
Automatic mask generation
Degraded image processing

Use Cases

Computer vision
Degraded image segmentation
Image segmentation under degraded conditions like blur, haze, and low light
Significant performance improvement compared to original SAM
Medical image analysis
Processing low-quality medical image segmentation
Autonomous driving
Scene understanding under adverse weather conditions
Image processing
Image dehazing
Serves as a preprocessing segmentation step for dehazing tasks
Improves downstream dehazing performance
Image deblurring
Serves as a preprocessing segmentation step for deblurring tasks
Improves downstream deblurring performance
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase