S2C2Seg：面向开放词表分割的语义-空间一致性与类别优化
S2C2Seg: Semantic-Spatial Consistency and Category Optimization for Open-Vocabulary Segmentation
免训练即插即用的类别子集筛选与置信度加权融合模块，可直接叠加到 SCLIP/ProxyCLIP/CorrCLIP 等 CLIP 密集预测方法上，缓解开放词表下的类别冗余与空间精度不足。
原文依据 S2C2Seg, a training-free framework that integrates with existing methods through Category Subset Selection (CSS) and Consistent Semantic Guidance (CSG).
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论开放词表语义分割将像素级识别扩展到任意文本描述的类别。尽管 CLIP 等视觉语言模型具备强大的全局语义理解能力，但其空间精度有限，且在大规模词表下存在语义歧义，制约了其在密集预测中的效果。本文提出 S2C2Seg，一个免训练框架，通过类别子集选择（CSS）与一致语义引导（CSG）与现有方法集成。CSS 采用三种互补的评分函数筛选类别候选：基于 CLIP 的全局语义相似度、来自密集预测模型的空间存在性，以及通过对齐与条件熵计算的多视图一致性。这种对语义、空间与一致性线索的联合利用降低了类别冗余与语义歧义。CSG 通过类别特定的置信度加权，自适应地融合 CLIP 全局特征与局部空间预测，对高相似度类别施加更强的正则化以纠正预测偏差，同时为低置信度类别保留空间精度。在八个基准上的大量实验证明了其广泛适用性：与 SCLIP、ProxyCLIP 和 CorrCLIP 集成时，S2C2Seg 在 mIoU 上取得 3.4 至 9.7 个百分点的稳定提升，以 51.2% 的平均 mIoU 创下新的最优结果。
英文原文摘要 1402 字符
Open-vocabulary semantic segmentation extends pixel-level recognition to arbitrary text-described categories. Despite strong global semantic understanding, vision-language models such as CLIP exhibit limited spatial precision and semantic ambiguity across large vocabularies, constraining their effectiveness for dense prediction. We present S2C2Seg, a training-free framework that integrates with existing methods through Category Subset Selection (CSS) and Consistent Semantic Guidance (CSG). CSS employs three complementary scoring functions to filter category candidates: CLIP-based global semantic similarity, spatial presence from dense prediction models, and multi-view consistency via alignment and conditional entropy. This joint exploitation of semantic, spatial, and consistency cues reduces category redundancy and semantic ambiguity. CSG adaptively fuses CLIP global features with local spatial predictions through category-specific confidence weighting, applying stronger regularization to high-similarity categories for correcting prediction biases while preserving spatial precision for low-confidence categories. Extensive experiments across eight benchmarks demonstrate broad applicability: when integrated with SCLIP, ProxyCLIP, and CorrCLIP, S2C2Seg achieves consistent improvements of 3.4 to 9.7 percentage points in mIoU, establishing a new state-of-the-art of 51.2% average mIoU.