Rethinking Multi-Branch and Cross-Backbone Fusion for Vehicle Re-Identification in the Foundation-Model Era

📄 arXiv: 2607.22068v1 📥 PDF

作者: Yu Wang, Hongyu Yang

分类: cs.CV, cs.LG

发布日期: 2026-07-24

💡 一句话要点

重新思考多分支与跨骨干融合以提升车辆重识别性能

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 车辆重识别 多分支架构 基础模型 性能提升 重排序 卷积神经网络 视觉Transformer

📋 核心要点

  1. 现有的多分支架构在车辆重识别中未能显著提升性能，且增加了计算复杂度。
  2. 论文提出通过优化单一强大的基础模型骨干和无训练重排序来提升车辆重识别效果。
  3. 实验结果显示，单一ConvNeXt模型在多个基准上表现优异，且多分支融合的性能提升有限。

📝 摘要（中文）

多分支架构和CNN-Transformer融合长期以来被认为是通过结合互补表示来提升车辆重识别（Re-ID）效果的有效方法。本研究在基础模型时代重新审视这一假设，通过全面的实证研究发现，单一的DINOv3预训练ConvNeXt模型在VeRi-Wild Small和VeRi-Wild Large数据集上分别达到了88.19和77.47的mAP，匹配了最强的多分支基线。通过无训练的重排序进一步提升性能至92.38和83.68 mAP。研究表明，增加表示多样性并未显著提升性能，而改进单一强大的基础模型骨干结合重排序阶段的策略更为有效。

🔬 方法详解

问题定义：本论文旨在解决车辆重识别中多分支架构和跨骨干融合的有效性问题。现有方法在提升性能的同时，增加了计算复杂度，未能带来显著的性能提升。

核心思路：论文的核心思路是通过优化单一的基础模型骨干（DINOv3预训练ConvNeXt）并结合无训练重排序，来验证在基础模型时代是否仍需依赖多分支架构。

技术框架：整体架构包括单一的ConvNeXt骨干网络，结合重排序模块进行性能提升。通过对比多分支架构与单一骨干的表现，评估其有效性。

关键创新：最重要的技术创新在于通过实证研究表明，单一强大的基础模型骨干结合重排序的策略比增加多分支或异构骨干更为有效。

关键设计：在实验中，使用了调优的训练方案，设置了合适的损失函数，并对网络结构进行了优化，以确保在不同数据集上达到最佳性能。

🖼️ 关键图片

fig_0
img_1
img_2

📊 实验亮点

实验结果显示，单一的DINOv3预训练ConvNeXt模型在VeRi-Wild Small和Large数据集上分别达到了88.19和77.47的mAP，经过无训练重排序后，性能进一步提升至92.38和83.68 mAP。与多分支基线相比，性能提升幅度有限，表明单一骨干模型的优化更为有效。

🎯 应用场景

该研究的潜在应用领域包括智能交通系统、监控视频分析和自动驾驶等。通过提升车辆重识别的准确性，可以有效增强交通管理和安全监控的智能化水平，未来可能对城市交通管理和公共安全产生深远影响。

📄 摘要（原文）

Multi-branch architectures and CNN-Transformer fusion have long been regarded as effective ways to improve vehicle re-identification (Re-ID) by combining complementary representations. In this work, we revisit this assumption in the foundation-model era through a comprehensive empirical study. A single DINOv3-pretrained ConvNeXt trained with a tuned recipe achieves 88.19 mAP on VeRi-Wild Small and 77.47 mAP on VeRi-Wild Large using visual cues alone, matching the strongest protocol-verified metadata-dependent multi-branch baseline. Applying training-free re-ranking further improves performance to 92.38 and 83.68 mAP, respectively. Using this strong baseline together with retrieval-level branch diagnostics, we evaluate whether increasing representational diversity still provides measurable gains. Across both benchmarks, concatenating multiple branches built on a shared backbone changes the best single-branch performance by less than one mAP point while increasing the embedding dimension by 4x, and the resulting representation has an effective rank close to the original feature dimension. We further study cross-backbone fusion using an asymmetric frozen-anchor strategy to combine ConvNeXt and Vision Transformer representations. Despite these favorable conditions, Transformer branches consistently remain 13-15 mAP below the ConvNeXt backbone, and paired per-query bootstrap analysis estimates the largest observed fusion gain to be only +0.11 mAP (95% confidence interval). Our results suggest that, under the evaluated setting, improving a single strong foundation-model backbone together with retrieval-stage re-ranking is more effective than increasing architectural complexity through additional branches or heterogeneous backbones. We restrict our conclusions to single-seed training and one family of foundation models and discuss conditions under which these observations may not hold.