针对车辆分类中的极端类别不平衡问题，比较了集成学习与CNN的性能表现。
Vehicle Classification under Extreme Imbalance: A Comparative Study of Ensemble Learning and CNNs
- 融合SMOTE与针对性欠采样构建平衡数据集，测试轻量级集成模型与可配置CNN
- CNN在全集上达79.19%准确率，未见批次达81.25%，优于集成方法的74.8%
- 少数类（驳船）仍识别失败，提示需结合成本敏感损失和增量采样
准确的车辆类型识别是智能交通与物流的基础，但公共数据集中严重的类别不平衡会抑制对稀有类别的性能。我们通过整合Kaggle、ImageNet及网络爬取数据，构建了一个包含16个类别的数据集（约4.7万张图像），并利用SMOTE过采样和目标性欠采样生成六种平衡版本。对比测试了随机森林、AdaBoost及基于MobileNet-V2特征的软投票组合器等轻量级集成模型，以及经过强增强和标签平滑训练的可配置ResNet-style CNN。最佳集成模型（SMOTE组合）在测试集上达到74.8%准确率，而CNN在完整测试集上取得79.19%准确率，在未见推理批次上达81.25%，验证了深度模型的优势。然而，最少数类（驳船）仍是失败模式，凸显仅靠重平衡的局限性。结果建议优先增加少数类样本采集，并采用代价敏感目标函数（如Focal Loss），或探索混合集成与CNN的流水线以兼顾可解释性与表征能力。
原文摘要 · Abstract (English)
Accurate vehicle type recognition underpins intelligent transportation and logistics, but severe class imbalance in public datasets suppresses performance on rare categories. We curate a 16-class corpus (~47k images) by merging Kaggle, ImageNet, and web-crawled data, and create six balanced variants via SMOTE oversampling and targeted undersampling. Lightweight ensembles, such as Random Forest, AdaBoost, and a soft-voting combiner built on MobileNet-V2 features are benchmarked against a configurable ResNet-style CNN trained with strong augmentation and label smoothing. The best ensemble (SMOTE-combined) attains 74.8% test accuracy, while the CNN achieves 79.19% on the full test set and 81.25% on an unseen inference batch, confirming the advantage of deep models. Nonetheless, the most under-represented class (Barge) remains a failure mode, highlighting the limits of rebalancing alone. Results suggest prioritizing additional minority-class collection and cost-sensitive objectives (e.g., focal loss) and exploring hybrid ensemble or CNN pipelines to combine interpretability with representational power.
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。