超轻量ViT模型，专为移动端图像分类设计，兼顾精度与效率。
UtVAA: Ultra-tiny Vision Transformer with Affix Attention for Mobile Image Classification
- 用局部-全局融合的附着注意力机制替代传统自注意力
- 最小模型仅20万参数、5400万次浮点运算，准确率接近主流模型
- 适合资源受限设备部署，尤其适用于移动和边缘场景
视觉变换器（ViTs）在图像分类中展现出强大表征能力，但其二次方复杂度和高参数量限制了在资源受限的移动端和边缘设备上的部署。本文提出超轻量视觉变换器UtVAA，专为严格计算预算下的高效视觉识别而设计。其核心是新型附着注意力（Affix Attention）模块，结合深度可分离局部特征提取、线性自注意力、坐标注意力建模空间依赖，并采用轻量三值融合策略整合局部与全局信息。此外，扩张瓶颈块通过扩张深度可分离卷积扩展感受野，同时保持低浮点运算量和稳定的优化。UtVAA分为可扩展的微型、中型和大型版本，最小模型含204.67K参数和53.95M FLOPs。在CIFAR-10、CIFAR-100、PlantVillage-Tomato和SLIF-Tomato数据集上的实验表明，该模型在百万参数以下的范围内实现了有竞争力的准确率。结果证明，基于Transformer的视觉模型可被重构为超轻量架构，且不显著损失判别性能，适用于移动和边缘部署。代码已开源。
原文摘要 · Abstract (English)
Vision Transformers (ViTs) have demonstrated strong representation capability in image classification. However, their quadratic self-attention complexity and large parameter counts limit deployment on resource-constrained mobile and edge devices. This paper introduces UtVAA, an ultra-tiny Vision Transformer architecture designed for efficient visual recognition under strict computational budgets. It incorporates a novel Affix Attention block that combines depthwise-pointwise local feature extraction, linear self-attention, coordinate attention for spatial dependency modelling, and a lightweight ternary fusion strategy to integrate local and global representations. In addition, Dilated Bottleneck blocks expand the receptive field using dilated depthwise separable convolutions while maintaining low FLOPs and stable optimisation through residual connections. UtVAA is implemented in scalable Tiny, Medium, and Large variants, with the smallest model containing 204.67K parameters and 53.95M FLOPs. Experimental results on CIFAR-10, CIFAR-100, PlantVillage-Tomato and SLIF-Tomato datasets show that UtVAA achieves competitive accuracy within a sub-million-parameter regime. Overall, the results demonstrate that transformer-based vision models can be redesigned into ultra-tiny architectures without significant loss in discriminative performance, making UtVAA suitable for mobile and edge deployment. Code is available at https://github.com/romiyal/UtVAA
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。