arXiv cs.AI· Haoyu Zheng, Jiang Liu, Jiaqi Zhu, Feifei Shao, Zheqi Lv, Wenqiao Zhang·· 9 天前AI 评分27
IAD-Unify：面向工业异常理解、分割与生成的任务专用接口
IAD-Unify: Task-Specific Interfaces for Industrial Anomaly Understanding, Segmentation, and Generation
AI 导读
IAD-Unify 通过任务专用 token 接口统一 MLLM、稠密视觉专家与扩散编辑器，实现工业异常的理解、分割与生成。其多参考 DINOv2 通路将 1,369 个单元压缩为 81 个证据 token 供 Qwen3.5 使用，另有 256-query 接口将 Qwen 状态重采样为 Stable Diffusion 的交叉注意力上下文。
正文 · 原文
Abstract:Industrial anomaly inspection requires complementary capabilities: explaining an observed defect, localizing its pixels, and synthesizing a controlled edit. We present IAD-Unify, a unified architecture connecting a multimodal language model (MLLM), dense visual expert, and diffusion editor through task-specific token interfaces. A multi-reference DINOv2 pathway forms a dense anomaly field and compresses its 1,369 cells into 81 structured evidence tokens. Qwen3.5 consumes these tokens for grounded answers and, with 32 task tokens, converts them into a semantic residual over the dense mask. A separate 256-query interface resamples Qwen states into Stable Diffusion's complete cross-attention context, while the editor retains its source latent and hard-mask inputs. The task pathways share one Qwen adaptation without forcing every task through the same visual bottleneck; in particular, evidence tokens are excluded from the generation pathway. Staged optimization initializes dense evidence, aligns it with language, calibrates segmentation, and then pretrains and specializes the diffusion editor while preserving earlier capabilities. We also construct Anomaly Evidence over 54,501 deduplicated industrial images. Its quality-controlled Anomaly Evidence Compiler and Industrial Edit-Pair Compiler produce family-disjoint, validated supervision through independent geometry, semantic-grounding, and edit-consistency checks. This design provides one fixed shared parameter set for understanding, segmentation, and localized generation without conflating their inputs, supervision, or outputs. The resulting model reaches 73.02% MMAD Macro$_7$, the highest listed public segmentation AP average (57.10%) with one reference, and the lowest Controlled masked DINO distance (0.3765), with complementary strengths across reasoning, pixel ranking, and semantic edit fidelity.
| Comments: | 16 pages, 11 figures. Revised title, author list, methodology, and experiments; supplementary material included |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2604.12440 [cs.CV] |
| (or arXiv:2604.12440v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2604.12440 arXiv-issued DOI via DataCite |
Submission history
From: Haoyu Zheng [view email]
[v1]
Tue, 14 Apr 2026 08:29:31 UTC (3,244 KB)
[v2]
Thu, 8 Oct 2026 09:03:20 UTC (2,074 KB)
来源：arXiv cs.AI · arxiv.org