轻量检测器通过知识蒸馏实现个性化目标识别，效果显著且推理成本低。
MOCHA: Multi-modal Objects-aware Cross-arcHitecture Alignment
- 用多模态教师模型蒸馏区域级知识，指导轻量学生模型训练
- 在四个基准上平均提升10.1%，少样本场景下表现优异
- 无需修改教师模型或推理时输入文本，适合移动端部署
个性化目标检测旨在仅用少量样本将通用检测器适配为识别用户特定实例。轻量模型因语义先验弱而表现不佳，而大型视觉语言模型（VLM）虽具备强对象理解能力，却因计算开销大难以用于实时或设备端应用。本文提出MOCHA（多模态对象感知跨架构对齐），一种知识蒸馏框架，将冻结的VLM教师模型中的多模态区域级知识迁移至轻量视觉模型。MOCHA提取融合的视觉与文本教师嵌入，通过双目标损失引导学生训练，强制区域间局部对齐与全局关系一致性。该过程实现高效语义迁移，无需修改教师模型或推理时输入文本。在严格少样本条件下，MOCHA在四个个性化检测基准上持续优于先前基线，平均提升10.1%，且推理开销极小。
原文摘要 · Abstract (English)
Personalized object detection aims to adapt a general-purpose detector to recognize user-specific instances from only a few examples. Lightweight models often struggle in this setting due to their weak semantic priors, while large vision-language models (VLMs) offer strong object-level understanding but are too computationally demanding for real-time or on-device applications. We introduce MOCHA (Multi-modal Objects-aware Cross-arcHitecture Alignment), a distillation framework that transfers multimodal region-level knowledge from a frozen VLM teacher into a lightweight vision-only detector. MOCHA extracts fused visual and textual teacher's embeddings and uses them to guide student training through a dual-objective loss that enforces accurate local alignment and global relational consistency across regions. This process enables efficient transfer of semantics without the need for teacher modifications or textual input at inference. MOCHA consistently outperforms prior baselines across four personalized detection benchmarks under strict few-shot regimes, yielding a +10.1 average improvement, with minimal inference cost.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。