我们能构建而非分类场景图吗？FlowSG：基于流匹配的渐进式图像条件场景图生成
Can We Build Scene Graphs, Not Classify Them? FlowSG: Progressive Image-Conditioned Scene Graph Generation with Flow Matching
把流匹配从图像生成迁移到结构化图输出：用连续速度场搬运框几何、用离散流更新类别 token，为视觉中「结构化预测当作生成任务」提供了可复用的混合离散-连续生成范式，且可插拔到现有检测器/分割器上。
原文依据 recasts SGG as continuous-time transport on a hybrid discrete-continuous state
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论场景图生成（SGG）通过预测边界框与主语-谓语-宾语三元组，统一了目标定位与视觉关系推理。然而多数流程仍将 SGG 视为一次性、确定性的分类任务，而非真正的渐进式生成任务。我们提出 FlowSG，将 SGG 重新表述为混合离散-连续状态上的连续时间传输：从含噪图出发，模型通过约束感知的逐步细化，联合合成节点（目标）与边（谓语），从而渐进地生长出图像条件场景图。具体而言，我们首先利用 VQ-VAE 将场景图（例如连续视觉特征）量化为紧凑、可预测的 token；随后由图 Transformer（i）预测条件速度场以传输连续几何（边界框），并（ii）更新类别 token（目标特征与谓语标签）的离散后验，通过流条件消息聚合将语义与几何耦合。训练将几何的流匹配损失与 token 的离散流目标相结合，实现少步推理，并可与标准检测器/分割器即插即用。在 VG 与 PSG 上于闭词表和开放词表协议下的大量实验表明，在谓语 R/mR 与图级指标上均取得一致提升，验证了混合离散-连续生成式表述优于一次性分类基线，例如相较 SOTA 的 USG-Par 平均提升约 3 个点。
英文原文摘要 1504 字符
Scene Graph Generation (SGG) unifies object localization and visual relationship reasoning by predicting boxes and subject-predicate-object triples. Yet most pipelines treat SGG as a one-shot, deterministic classification instead of a genuine progressive, generative task. We propose FlowSG, which recasts SGG as continuous-time transport on a hybrid discrete-continuous state: starting from a noised graph , the model progressively grows an image-conditioned scene graph through constraint-aware refinements that jointly synthesize nodes (objects) and edges (predicates). Specifically, we first leverage a VQ-VAE to quantize a scene graph (e.g., the continuous visual features) into compact, predictable tokens; a graph Transformer then (i) predicts a conditional velocity field to transport continuous geometry (boxes) and (ii) updates discrete posteriors for categorical tokens (object features and predicate labels), coupling semantics and geometry via flow-conditioned message aggregation. Training combines flow-matching losses for geometry with a discrete-flow objective for tokens, yielding few-step inference and plug-and-play compatibility with standard detectors/segmenters. Extensive experiments on VG and PSG under closed- and open-vocabulary protocols show consistent gains in predicate R/mR and graph-level metrics, validating the mixed discrete-continuous generative formulation over one-shot classification baselines, e.g., an average improvement of about 3 points over the SOTA USG-Par.