Tensorflow for IoT

Tensorflow for IoT

• 1.
  Tensorflow for IoT GEETACHAUHAN JUNE, 2017
• 2.
  Agenda  Era ofAI First: What does it mean for IoT?  What is Deep Learning?  Use Cases for IoT  Tensorflow for IoT  Optimizations for IoT  Common Problem Solutions  Look into the Future  References
• 3.
  Era of AIFirst  Billions of connected devices  Intelligence at the Edge  Increasing Computation power  Edison: 500 MHz, 1 GB RAM  RPi3: 1.2 GHz Quad-core  Deep neural networks running on the IoT device  Local inferencing → compressed insights to cloud
• 4.
  What is Deep Learning?  AINeural Networks composed of many layers  Learn like humans  Automated Feature Learning  Layers are like Image Filters
• 5.
  Use Cases for IoT Cucumber sorter  Real time arrival prediction for Caltrains  Wearable assistant for Blind (Horus)  Smart cameras, smart door lock  Human line counter  Real time exercise score on smart watch  Creative Arts, Music Generators  Self driving cars, Intelligent Robots
• 6.
  Cucumber Sorter
• 7.
  Caltrain Rider  RealtimeCaltrain arrival prediction  Audio Visual pipeline on Raspberry pi  Image classification for Caltrains
• 8.
  Tensorflow for IoT  Compilefrom source  Android Things Tensorflow Inference Library (in preview now)  Tensorflow Lite (Announced @Google IO for later this year)
• 9.
  Tensorflow IoT Pipeline Inferenceon DeviceTrain Model in Cloud
• 10.
  Tensorflow Android Things  AddTensorflow Inferencing library  Create Inference object, load model  feed() image, run() to predict  fetch() to get classification result
• 11.
  Demo Intel Edison +Seed Studio Kit Raspberry Pi + Pi Camera
• 12.
  Transfer Learning Transfer Learning –New Classifier Inception V3
• 13.
  Optimizations  Graph TransformTool  Freeze graph (variables to constants)  Quantize weights (20 M weights for IV3)  Quantization (32 bit float → 8 bit float)  Memory Mapping  Inception v3 93 MB → 1.5 MB
• 14.
  Common Problem Solutions Tensorboard is your friend – X-Ray vision  Image size mismatch for the input tensor  Output classifier specify correct number of classes from your model  Model too large – Load in memory  Missing ops -- Graph transform tool  Device heating up under heavy processing load  Split the model, do part detection on device, rest in cloud  Reduce the frequency eg only do on movement detection
• 15.
  Look into theFuture  Hardware: Neural Network Chips  Software: Tensorflow Lite (Android NN) Intel Fathom Neural Stick Nvidia Jetson
• 16.
  References  Tensorflow forAndroid Things Sample  Tensorboard hands-on  Graph Transform Tool  Tensorflow for Poets (Transfer learning)  Tensorflow for Poets2 (Optimizations for Mobile/IoT)  Cucumber Farmer Deep Learning Story  Caltrain Rider story  Intel Fathom Neural Stick  Nvidia Jetson
• 17.
  Questions? Contact https://www.linkedin.com/ in/geetachauhan/ geeta@svsg.co