Scholarlink

<p>Making decisions in real time on edge devices is a must for autonomous cars, smart clinics, and factory IoT, yet the squeeze of limited compute power, high speed requirements, and another shortage of labeled examples still holds back the field. Classic deep-learning frameworks hinge on big centralized GPUs and huge supervised datasets, keeping them off the edge. With self-supervised, cross-modal transformers, the situation changes. These lighter-weight models exploit correlations across different data types without full annotation and adapt on the spot to shifting inputs. The result is a decision-making backbone that is resilient, quick, and light on power—all while operating natively on edge hardware.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Self-Supervised Multi-modal Transformers for Real-Time Decision Making in Edge AI Systems

  • 期刊论文

  • 发表日期：

    2026年5月2日

查看全文
  • Shubhkirti Bodkhe

摘要

Making decisions in real time on edge devices is a must for autonomous cars, smart clinics, and factory IoT, yet the squeeze of limited compute power, high speed requirements, and another shortage of labeled examples still holds back the field. Classic deep-learning frameworks hinge on big centralized GPUs and huge supervised datasets, keeping them off the edge. With self-supervised, cross-modal transformers, the situation changes. These lighter-weight models exploit correlations across different data types without full annotation and adapt on the spot to shifting inputs. The result is a decision-making backbone that is resilient, quick, and light on power—all while operating natively on edge hardware.