Grounded SAM: Assembling Open-World Models for Diverse Visual Tasks

TL;DR

Grounded SAM combines open-set object detection with segmentation to enable versatile, open-world visual tasks, integrating various models for annotation, editing, and 3D analysis, and demonstrating strong zero-shot performance.

Contribution

The paper introduces Grounded SAM, a novel framework that integrates Grounding DINO with SAM for open-world visual understanding and task versatility.

Findings

Achieves 48.7 mean AP on SegInW zero-shot benchmark.

Enables diverse tasks like annotation, image editing, and 3D motion analysis.

Demonstrates superior open-vocabulary segmentation performance.

Abstract

We introduce Grounded SAM, which uses Grounding DINO as an open-set object detector to combine with the segment anything model (SAM). This integration enables the detection and segmentation of any regions based on arbitrary text inputs and opens a door to connecting various vision models. As shown in Fig.1, a wide range of vision tasks can be achieved by using the versatile Grounded SAM pipeline. For example, an automatic annotation pipeline based solely on input images can be realized by incorporating models such as BLIP and Recognize Anything. Additionally, incorporating Stable-Diffusion allows for controllable image editing, while the integration of OSX facilitates promptable 3D human motion analysis. Grounded SAM also shows superior performance on open-vocabulary benchmarks, achieving 48.7 mean AP on SegInW (Segmentation in the wild) zero-shot benchmark with the combination of Grounding DINO-Base and SAM-Huge models.