MicroViTv2优化边缘设备视觉模型，提升精度与能效。
MicroViTv2: Beyond the FLOPS for Edge Energy-Friendly Vision Transformers
- 采用重参数化结构，加速推理并减少冗余。
- 在ImageNet-1K上精度提升0.5%，超越MobileViTv2等模型。
- 适合低功耗边缘设备部署，关注真实硬件效率。
视觉变压器（ViT）在各类视觉任务中表现优异，但计算开销大，难以在边缘设备部署。本文提出轻量级的MicroViTv2，基于原始MicroViT改进，引入重参数化设计：重参数化补丁嵌入（RepEmbed）和重参数化深度卷积混合器（RepDW），实现更快推理；同时提出单深度转置注意力（SDTA），以最小冗余捕捉长程依赖。尽管浮点运算量（FLOPs）略有增加，但模型在精度上比前代提升0.5%，优于MobileViTv2、EdgeNeXt和EfficientViT。在Jetson AGX Orin上验证，兼具快速推理与高能效。图像分类（ImageNet-1K）与目标检测（COCO）实验表明，硬件感知设计与结构重参数化是实现高精度与低能耗的关键，强调评估效率需超越单纯FLOPs。代码已开源。
原文摘要 · Abstract (English)
The Vision Transformer (ViT) achieves remarkable accuracy across visual tasks but remains computationally expensive for edge deployment. This paper presents MicroViTv2, a lightweight Vision Transformer optimized for real-device efficiency. Built upon the original MicroViT, the proposed model is designed based on reparameterized design, specifically Reparameterized Patch Embedding (RepEmbed) and Reparameterized Depth-Wise convolution mixer (RepDW) for faster inference, and introduces the Single Depth-Wise Transposed Attention (SDTA) to capture long-range dependencies with minimal redundancy. Despite slightly higher FLOPs, MicroViTv2 improves accuracy up to 0.5% compared to its predecessor and surpassing MobileViTv2, EdgeNeXt, and EfficientViT while maintaining fast inference and high energy efficiency on Jetson AGX Orin. Experiments on ImageNet-1K and COCO demonstrate that hardware-aware design and structural re-parameterization are key to achieving high accuracy and low energy consumption, validating the need to evaluate efficiency beyond FLOPs. Code is available at https://github.com/novendrastywn/MicroViT.
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。