How to Train Your Dragon and Its Friends: AI on the Edge with Eclipse Kura
summary
A complete edge ML pipeline built end-to-end: sensor data from a Raspberry Pi + Sensat board is published with Kura Wires to Eclipse Kapua, exported as CSV, trained in TensorFlow/Keras as an autoencoder for anomaly detection, converted into a Triton-compatible ensemble (preprocessor, model, postprocessor) and run in a Dockerized Nvidia Triton™ Server on the Pi. Reconstruction error threshold detects injected anomalies, deployed model reports anomalies back to Kapua, model encryption and additional inference backends are supported, and the demo works live on-device.
AI-written summary · about
chapters (7)
- 0:00Start
- 11:52DATA COLLECTION
- 12:39PUBLISHER CONFIGURATION
- 13:24DRIVER CONFIGURATION
- 16:02MODEL BUILDING AND TRAINING
- 27:23MODEL DEPLOYMENT
- 27:44TRITON COMPONENT CONFIGURATION
related talks
55mLarge Scale Distributed AI on the JVM
31mIoT hardware access made easy with Eclipse Kura
29mAnomaly Detection
10mReady, Set, Go for K8s at edge today for AI/ML with Tensorflow
36mApplied AI with Eclipse Graphene
45mAIoT Ops - Using Kubernetes and ML Ops to Build Edge ML Applications
01h 02mAI on the Edge
31mPrecision farming powered by K3s and TensorRT