Segmentation model separates meaning and location for better masks
SeGDeP: Semantic- and Geometric-Aware Decoupled Prompts for Reasoning Segmentation
Computer Vision and Pattern Recognition
Summary
Identifying the exact area that a language description refers to in images is hard because it needs understanding what something is and where it is separately. The authors created a method that handles the meaning and the location parts independently before combining them, making it easier to pinpoint the right area. Their approach improves accuracy by using separate paths for semantic meaning and geometric location to guide the image mask creation. This helps machines better interpret instructions to highlight image regions precisely.
What this means in practice
- •For computer vision engineers: Build image segmentation tools that separately process meaning and location to improve segmentation accuracy from natural language queries.
- •For augmented reality developers: Create AR systems that better identify objects from spoken or written descriptions by separating semantic understanding from spatial location.
Authors
Linnan Zhao, Xu Liu, Lingling Li, Licheng Jiao, Fang Liu, Wenping Ma
Abstract
Reasoning segmentation converts an implicit linguistic conclusion into a precise mask, requiring both semantic identification and spatial grounding. Existing MLLM-segmenter interfaces either use a special trigger or compress both signals into one context, although they receive different supervision and fail differently. This coupling obscures whether a failure arises from target interpretation or from localization. We present SeGDeP, an explicit what-where interface. A semantic prompt branch and an independent geometric projection path transform resolved MLLM states into semantic features and a DETR-predicted box, which jointly condition a SAM 3 mask decoder. Training first aligns this executable interface, then uses group reward-decoupled policy optimization (GDPO) to balance format, box-IoU, and mask-IoU feedback. SeGDeP-4B reaches 82.7 average cIoU over eight RefCOCO-family splits and 66.0/59.6 gIoU on ReasonSeg val/test while adapting only 0.38% of Qwen3-VL parameters through LoRA. Controlled stage-wise ablations, gradient diagnostics, and prompt interventions further show that the two paths develop complementary semantic and geometric specialization rather than duplicating the same evidence.