Using TensorFlow for Performing Deep Learning at the Edge
summary
Claims that edge ML must scale to ~21 billion IoT devices by 2025 and shows concrete toolchains: TensorFlow.js + Node.js for browser/client online training and inference (transfer learning demo controlling a Pac‑Man agent), TensorFlow Lite for quantized offline models deployed to Raspberry Pi, ESP32 and Jetson (8‑/16‑bit quantization via the TFLite converter), and TensorFlow Hub for registering/versioning models and CI/CD-style pulls. Includes a working Raspberry Pi object‑detection demo that runs in real time.
AI-written summary · about
chapters (15)
- 0:00Introduction to edge ML
- 2:50Why machine learning at the edge
- 7:06Cloud-based ML workflow
- 9:14TensorFlowJS overview
- 12:05Online vs offline training
- 14:51Demo of transfer learning
- 17:41TensorFlowJS demos and Node Red
- 20:55Use cases for real-time ML
- 24:25TensorFlow Lite introduction
- 30:50TensorFlow Lite demo
- 33:51Supported edge devices
- 36:42TensorFlow Hub overview
- 41:59TensorFlow Hub demos
- 44:27Online vs offline ML summary
- 46:32Q&A session
related talks
31mDeep Learning at the Edge with TensorFlow.js
01h 01mML and the IoT: Living on the Edge
37mDeep Learning Neural Network Acceleration at the Edge
49mEnable IoT with Edge Computing and Machine Learning
27mMachine Learning on the edge using TensorFlow Lite
55mEdgeAI - AI in the Palm of Your Hand
20mTensorFlow.js Bringing Machine Learning to the Web and Beyond
28mMachine Learning on the Edge