跳转至

FaceLiVTv2: 移动端轻量人脸识别

type: concept tags: [mobile, face-recognition, hybrid-architecture, cnn-transformer, edge, 其他] related: [[fastshade-mobile-denoising]], [[on-device-inference-memory-pressure]], [[coremltools-9]] sources: - http://arxiv.org/abs/2604.09127v1 created: 2026-04-14

核心问题

This paper introduces FaceLiVT, a lightweight yet powerful face recognition model that integrates a hybrid Convolution Neural Network (CNN)-Transformer architecture with an innovative and lightweight Multi-Head Linear Attention (MHLA) mechanism. By combining MHLA alongside a reparameterized token mixer, FaceLiVT effectively reduces computational complexity while preserving competitive accuracy. Extensive evaluations on challenging benchmarks; including LFW, CFP-FP, AgeDB-30, IJB-B, and IJB-C; highlight its superior performance compared to state-of-the-art lightweight models. MHLA notably impro

论文信息

  • 标题: FaceLiVT: Face Recognition using Linear Vision Transformer with Structural Reparameterization For Mobile Device
  • 作者: Novendra Setyawan, Chi-Chia Sun, Mao-Hsiu Hsu
  • 来源: arXiv

方法/架构

详细方法论待补充。参考原始论文获取完整技术细节。

为什么重要

作为手机端 AIOS 生态的一部分，FaceLiVTv2：移动端高效人脸识别 对推动端侧 AI 落地具有重要意义。

关联

  • [[clawmobile-agentic]] — Agent 系统架构
  • [[kv-cache-quantization-ondevice]] — 内存优化