构建合成数据集，精准评估视觉模型是否真能推断部件-整体层级关系。
SynDaCaTE: A Synthetic Dataset For Evaluating Part-Whole Hierarchical Inference
- 设计新数据集SynDaCaTE，专用于测试胶囊网络的层级推理能力。
- 发现主流胶囊模型在部件组合推理上存在明确瓶颈。
- 证明自注意力机制对部件到整体推理效果极佳，适合研究视觉归纳偏置。
理解物体表征，特别是部件-整体层级关系，是计算机视觉中的核心课题，旨在提升数据效率、系统性泛化能力和鲁棒性。尽管被称为胶囊网络的模型被设计为能推断部件-整体层级，但它们通常在图像分类等监督任务中端到端训练，难以验证其是否真正学习了该能力。为此，我们提出了一个名为SynDaCaTE（SYNthetic DAtaset for CApsule Testing and Evaluation）的合成数据集，并通过两项实验验证其有效性：(1) 精准定位了主流胶囊模型在层级推理中的性能瓶颈；(2) 证明了置换等变的自注意力机制在从部件推断整体方面极为有效，为未来设计有效的视觉归纳偏置提供了新方向。
原文摘要 · Abstract (English)
Learning to infer object representations, and in particular part-whole hierarchies, has been the focus of extensive research in computer vision, in pursuit of improving data efficiency, systematic generalisation, and robustness. Models which are \emph{designed} to infer part-whole hierarchies, often referred to as capsule networks, are typically trained end-to-end on supervised tasks such as object classification, in which case it is difficult to evaluate whether such a model \emph{actually} learns to infer part-whole hierarchies, as claimed. To address this difficulty, we present a SYNthetic DAtaset for CApsule Testing and Evaluation, abbreviated as SynDaCaTE, and establish its utility by (1) demonstrating the precise bottleneck in a prominent existing capsule model, and (2) demonstrating that permutation-equivariant self-attention is highly effective for parts-to-wholes inference, which motivates future directions for designing effective inductive biases for computer vision.
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。