FlowDIS：基于流匹配的语言引导二值图像分割
FlowDIS: Language-Guided Dichotomous Image Segmentation with Flow Matching
把流匹配的时变向量场用于从图像分布到掩码分布的传输，为高精度分割提供了一条不依赖判别式解码器的新范式；其位置感知实例配对训练策略与文本条件控制可直接迁移到开放词表分割、交互式分割等需要像素级可控性的任务。
原文依据 learns a time-dependent vector field to transport the image distribution to the corresponding mask distribution
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论精确的图像分割对于图像编辑、自动驾驶和医学图像分析等现代计算机视觉应用至关重要。近年来，二值图像分割（DIS）已成为训练和评估高精度分割模型的标准任务。现有的 DIS 方法往往无法保留细粒度细节，或无法充分捕捉前景的语义结构。为应对这些挑战，我们提出 FlowDIS，一种基于流匹配框架的新型二值图像分割方法，它学习一个时变向量场，将图像分布传输到相应的掩码分布，并可选地以文本提示为条件。此外，借助我们提出的位置感知实例配对（PAIP）训练策略，FlowDIS 通过文本提示提供了强大的可控性，从而实现精确的像素级目标分割。大量实验表明，无论是否使用语言引导，我们的方法都显著优于最优方法。与最优的已有 DIS 方法相比，FlowDIS 在 DIS-TE 测试集上取得了高 5.5% 的 F_beta^omega 指标和低 43% 的 MAE (M)。代码见：https://github.com/Picsart-AI-Research/FlowDIS。
英文原文摘要 1263 字符
Accurate image segmentation is essential for modern computer vision applications such as image editing, autonomous driving, and medical image analysis. In recent years, Dichotomous Image Segmentation (DIS) has become a standard task for training and evaluating highly accurate segmentation models. Existing DIS approaches often fail to preserve fine-grained details or fully capture the semantic structure of the foreground. To address these challenges, we present FlowDIS, a novel dichotomous image segmentation method built on the flow matching framework, which learns a time-dependent vector field to transport the image distribution to the corresponding mask distribution, optionally conditioned on a text prompt. Moreover, with our Position-Aware Instance Pairing (PAIP) training strategy, FlowDIS offers strong controllability through text prompts, enabling precise, pixel-level object segmentation. Extensive experiments demonstrate that our method significantly outperforms state-of-the-art approaches both with and without language guidance. Compared with the best prior DIS method, FlowDIS achieves a 5.5% higher F_beta^omega measure and 43% lower MAE (M) on the DIS-TE test set. The code is available at: https://github.com/Picsart-AI-Research/FlowDIS.