✨ X2SAM ✨

Any Segmentation in Images and Videos

Hao Wang1,2 Limeng Qiao3 Chi Zhang3 Guanglu Wan3 Lin Ma3 Xiangyuan Lan2📧 Xiaodan Liang1,2📧
1 Sun Yat-sen University 2 Peng Cheng Laboratory 3 Meituan Inc.
📧 Corresponding author
arXiv Paper Demo HuggingFace Code

Highlights

X2SAM introduces a unified multimodal large language model (MLLM) framework, extending the segmentation paradigm from segment anything to any segmentation, thereby enhancing pixel-level perceptual understanding.
X2SAM proposes a novel Visual GrounDed (VGD) segmentation task, which segments all instance objects using interactive visual prompts, empowering the model with visually grounded, pixel-wise interpretative capabilities.
X2SAM presents a unified training strategy enabling co-training across multiple datasets and achieves state-of-the-art performance on various image segmentation benchmarks with a single model.

Abstract

Large Language Models (LLMs) demonstrate strong capabilities in broad knowledge representation, yet they are inherently deficient in pixel-level perceptual understanding. Although the Segment Anything Model (SAM) represents a significant advancement in visual-prompt-driven image segmentation, it exhibits notable limitations in multi-mask prediction and category-specific segmentation tasks, and it cannot integrate all segmentation tasks within a unified model architecture.

To address these limitations, we present X2SAM, a streamlined Multimodal Large Language Model (MLLM) framework that extends the segmentation paradigm from segment anything to any segmentation. Specifically, we introduce a novel unified framework that enables more advanced pixel-level perceptual comprehension for MLLMs. Furthermore, we propose a new segmentation task, termed Visual GrounDed (VGD) segmentation, which segments all instance objects with interactive visual prompts and empowers MLLMs with visual grounded, pixel-wise interpretative capabilities. To enable effective training on diverse data sources, we present a unified training strategy that supports co-training across multiple datasets. Experimental results demonstrate that X2SAM achieves state-of-the-art performance on a wide range of image segmentation benchmarks, highlighting its efficiency for multimodal, pixel-level visual understanding.

Overview

X2SAM Framework Architecture
Fig. 1. The Overview of X2SAM. X2SAM comprises dual encoders, dual projectors, a language model, a segmentation connector, and a segmentation decoder. The dual encoders process the image and project features to match text embedding dimensions, which are then input to the language model with tokenized text for instruction-guided understanding. The SAM features are connected to the segmentation decoder, which uses the LLM's <SEG> token to generate segmentation masks.

Benchmarks

Tab. 1. Comprehensive Performance Comparison. We compare X2SAM to segmentation-specific models (Gray) and MLLMs. ✗ denotes unsupported tasks. "-" indicates unreported results. Best results are in bold, second-best are underlined.
Benchmark Results Overview
View Detailed Results
Tab. 2. Comparison of Referring Segmentation. Evaluated by (M)LLMs.
Referring Segmentation
Tab. 3. Comparison of GCG Segmentation. † indicates pretraining with the GranD dataset.
GCG Segmentation
Tab. 4. Comparison of VGD Segmentation. † indicates evaluation results following X2SAM setting.
VGD Segmentation
Tab. 5. Generic Segmentation.
Generic Segmentation
Tab. 6. OV Segmentation (A150-OV).
OV Segmentation
Tab. 7. Comparison of Reasoning Segmentation.
Reasoning Segmentation
Tab. 8. Comparison of Interactive Segmentation.
Interactive Segmentation
Tab. 9. Image-level Benchmarks. MME, MMBench, SEED-Bench, POPE, and AI2D.
Image-level Benchmarks

Live Demo

Launch Interactive Demo
0:00 0:00

Citation

@article{wang2025x2sam,
  title={X2SAM: Any Segmentation in Images and Videos},
  author={Wang, Hao and Qiao, Limeng and Zhang, Chi and Wan, Guanglu and Ma, Lin and Lan, Xiangyuan and Liang, Xiaodan},
  journal={arXiv preprint arXiv:2603.00000},
  year={2026}
}