From Ground to Space: Cloud-Native Edge Machine-Learning Case Studies with KubeEdge-Sedna
summary
Demos show KubeEdge-Sedna and Janus running lifelong learning on delivery robots and joint inference between satellites and ground: robots upload eight unseen images, cloud retrains via lifelong learning and reduces a repeat route’s delivery time by 28.05%, while satellite small-model / cloud big-model joint inference cuts transmission and energy needs and reports precision and recall up to 99%. Architecture details include async edge testing, unseen-task chaining, and cloud-coordinated model updates and scheduling.
AI-written summary · about
chapters (19)
- 0:00Introduction
- 0:21About me
- 1:07Agenda
- 1:35Edge Computing
- 3:18KubeEdge
- 5:30Discovery
- 7:54Distributed Collaborative AI
- 10:00Business Value
- 11:29Challenges
- 13:18Smart Robotic Perception
- 15:27Data Heterogeneity
- 16:51Edge Data Deficiency
- 17:33Lifelong Learning
- 18:34Edge Cloud Collaborative Learning
- 19:58Smart Robotic Perception Demo
- 26:01Smart Robotic Lifelong Learning Demo
- 29:37Benefits
- 30:04Problems
- 30:39Conclusion
related talks
15mKeynote: Incremental Deep Learning For Satellite with KubeEdge...
20mUsing Training-Operator To Schedule Distributed Edge-Cloud Collaborative...
25mManage Cloud Native LLM Workloads Across Edge and Cloud Seamlessly Using KubeEdge and WasmEdge
26mBuilding a Cloud Native Edge Platform for Accelerating AI Inference
31mBuild Your Intelligent Edge
49mEnable IoT with Edge Computing and Machine Learning
06mEdge AI in Practice: Lessons from Producti...
28mMachine Learning on the Edge