arXivDaily arXiv每日学术速递 周一至周五更新
​通知:联邦节假日调休:哥伦布日 / 原住民日(Columbus Day / Indigenous Peoples' Day),今日(13日)无更新。
arXiv 2602.09701cs.CVcs.AI

GenSeg-R1: 通过强化学习驱动的视觉-语言接地进行细粒度指称分割

GenSeg-R1: RL-Driven Vision-Language Grounding for Fine-Grained Referring Segmentation

Camcom Technologies Pvt. Ltd.(Camcom技术有限公司)

Sandesh Hegde, Jaison Saji Chacko, Debarshi Banerjee, Uma Mahesh

AI总结:

GenSeg-R1 通过强化学习驱动的视觉-语言接地方法,在细粒度指称分割任务中实现了优于基线的性能。

摘要:

我们研究了通过解耦的推理-分割管道进行细粒度指称图像分割。一个视觉-语言模型(VLM)接收一张图像和一个自然语言查询,对场景进行推理,并生成结构化的空间提示:每个指称实例都包含一个边界框和两个内部关键点。一个冻结的可提示分割器(SAM 2)将这些提示转换为高质量的掩码。在我们的 GenSeg-R1 框架中,我们使用 Group Relative Policy Optimization(GRPO)微调 Qwen3-VL 模型(4B 和 8B 参数),无需监督的推理链标注。在 RefCOCOg 验证集上,我们最好的模型(GenSeg-R1-8B)在 cIoU 上达到 0.7127,在 mIoU 上达到 0.7382,显著优于对应的 Qwen3-VL Instruct 基线(分别提高 15.3 和 21.9 分)。我们进一步引入了 GenSeg-R1-G,该模型在 GRefCOCO 上进行训练,使用带有 SAM 2 的奖励机制,直接优化掩码质量。在 GRefCOCO 验证集上,GenSeg-R1-G 在目标 mIoU 上达到 76.69%,在负样本(无目标)提示上的准确率高达 82.40%,显著优于 Seg-R1-7B 和 Seg-Zero-7B,后者缺乏无目标检测能力。在 ReasonSeg 测试中,GenSeg-R1-4B 达到 68.40% 的 mIoU,超过 Seg-Zero-7B 7.0 分,超过 Seg-R1-7B 10.7 分。

英文摘要:

We study fine-grained referring image segmentation via a decoupled reason-then-segment pipeline. A vision-language model (VLM) receives an image and a natural-language query, reasons about the scene, and emits structured spatial prompts: a bounding box plus two interior keypoints for every referred instance. A frozen promptable segmenter (SAM 2) converts these prompts into high-quality masks. Within our GenSeg-R1 framework we finetune Qwen3-VL models (4B and 8B parameters) using Group Relative Policy Optimization (GRPO), requiring no supervised reasoning-chain annotations. On RefCOCOg validation our best model (GenSeg-R1-8B) achieves 0.7127 cIoU and 0.7382 mIoU, substantially outperforming the corresponding Qwen3-VL Instruct baselines (+15.3 and +21.9 points, respectively) and surpassing Seg-Zero-7B [3] by +3.3 cIoU under identical evaluation. We further introduce GenSeg-R1-G, a variant trained on GRefCOCO [9] with a SAM 2 in-the-loop reward that directly optimizes mask quality. On GRefCOCO validation GenSeg-R1-G achieves 76.69% target mIoU with 82.40% accuracy on negative (no-target) prompts, substantially outperforming Seg-R1-7B and Seg-Zero-7B, which lack no-target detection capability. On ReasonSeg test, GenSeg-R1-4B reaches 68.40% mIoU, surpassing Seg-Zero-7B by +7.0 and Seg-R1-7B by +10.7 points.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑