---
格式版本: 2
标题: "FD-CanKD: Frequency-Decoupled Cross-Attention Distillation as a Refinement Prior for Compact Object Detectors"
原文链接: "https://arxiv.org/abs/2608.18590"
发布日期: "2026-08-19"
发布时间校准状态: "found"
发布时间需复核: "否"
发布时间来源: "rule:local:strict_original_body"
发布时间证据: "**\\[v1\\]** Wed, 19 Aug 2026 06:32:11 UTC (25,718 KB)"
发布时间校准原因: "规则确认唯一严格发布时间，来源 local:strict_original_body"
发布时间校准置信度: "high"
发布时间候选数量: 18
发布时间严格候选数量: 6
发布时间原页读取状态: "source template page reused from URL open"
发布时间未找到原因: ""
发布时间校准时间: "2026-08-20T14:39:35+08:00"
发布时间仲裁状态: "skipped"
发布时间仲裁尝试次数: 0
发布时间仲裁耗时毫秒: 0
发现时间: "2026-08-20T14:38:28+08:00"
入库时间: "2026-08-20T06:39:36.057Z"
来源平台: "arXiv 学术论文搜索"
搜索渠道: "source_template"
搜索词: "https://arxiv.org/search/?query=Microsoft&searchtype=all"
匹配关键词:
  []
相关厂家:
  - "Microsoft"
相关专家:
  []
内容类型: "网页"
抓取工具: "Free Fetch + Defuddle"
清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取"
原始附件:
  []
AI优质: "否"
AI打分: 5
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "论文摘要涉及物体检测知识蒸馏，与超节点/AI Rack/机柜级AI基础设施完全无关，属于明显无关内容。"
AI质检模型: "tx-deepseek-v4-flash"
AI质检时间: "2026-08-20T14:40:26+08:00"
AI主题相关性: 0
AI来源权威性: 5
AI新颖性: 0
AI技术细节: 0
AI商业部署信号: 0
AI完整性: 0
AI摘要: "FD-CanKD 提出一种面向紧凑目标检测器的频率解耦交叉注意力知识蒸馏框架，在预测、关系与频率三个层级迁移教师知识。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-07T03:18:38.630Z"
采集批次: "2026年8月20日14点19分32秒"
采集批次ID: "20260820-141932-079"
去重键: "https://arxiv.org/abs/2608.18590"
---

## Computer Science > Computer Vision and Pattern Recognition

## Title:FD-CanKD: Frequency-Decoupled Cross-Attention Distillation as a Refinement Prior for Compact Object Detectors

Authors:[YoungJae Cheong](https://arxiv.org/search/cs?searchtype=author&query=Cheong,+Y), [Jhonghyun An](https://arxiv.org/search/cs?searchtype=author&query=An,+J)

[View PDF](https://arxiv.org/pdf/2608.18590) [HTML (experimental)](https://arxiv.org/html/2608.18590v1)

> Abstract:Compact object detectors are suitable for resource-constrained visual perception, but their limited representation capacity creates an accuracy gap relative to large models. Conventional detector distillation often relies on prediction-level supervision or a single feature-alignment target, such as response, distribution, correlation, or frequency-domain matching. Frequency-Decoupled Cross-Attention Knowledge Distillation (FD-CanKD) is presented as a detector-oriented framework that transfers teacher knowledge at three complementary levels: head-level prediction supervision, relation-level non-local context transfer, and frequency-level component-selective alignment. Student features first aggregate teacher-side spatial context through cross-attention-based relation transfer, after which frequency-aware alignment preserves complementary structural and detail-sensitive cues. Under controlled Microsoft Common Objects in Context (COCO) experiments, fixed 50-epoch from-scratch comparisons show that FD-CanKD remains competitive with representative detector knowledge distillation baselines. Post-distillation continued fine-tuning further produces a stronger refinement-ready student than detector-only fine-tuning, reaching 48.87 mean average precision (mAP) at intersection-over-union thresholds from 0.50 to 0.95 (mAP50:95), 65.84 mAP50, and 53.40 mAP75 after 20 additional epochs. All distillation modules are removed after training, leaving the deployed student unchanged at 19.7M parameters. The framework is instantiated and evaluated in a controlled YOLOv12 teacher-student setting as a representative compact-detector case study.

| Comments: |  |
| --- | --- |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | [arXiv:2608.18590](https://arxiv.org/abs/2608.18590) \[cs.CV\] |
|  | (or [arXiv:2608.18590v1](https://arxiv.org/abs/2608.18590v1) \[cs.CV\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.18590](https://doi.org/10.48550/arXiv.2608.18590) |

## Submission history

From: Cheong YoungJae \[[view email](https://arxiv.org/show-email/62ad9219/2608.18590)\]  
**\[v1\]** Wed, 19 Aug 2026 06:32:11 UTC (25,718 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.18590) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))