ML on the edge

This is a presentation on how mobile applications can benefit by doing ML on the device itself as opposed to in the cloud. MLKit is a library introduced by Google for this purpose. This presentation also provides brief code examples of integrating MLKit in Android application.

Transcript

1. ML on the edge Integration of MLKit on mobile applications 1 WTM - WiDS Lightning talks Sangeetha KP (@sangee_kp)
2. Why run ML on edge? 1. Latency 2. Privacy of data 3. Real-time 4. Data usage 5. Offline 6. Experimentation and many more! 2
3. MLKit 1. Google’s machine learning SDK optimized for mobile applications 2. Tailored for your app (Android, iOS) 3. Built on top of TensorFlow lite 3
8. Implementation 1. Connect your app with firebase 2. Add Gradle MLKit dependencies 3. Run your image through the recognizer for processing 4. Retrieve and parse the processed results 8
9. Steps to integrate a MLKit library 1. Connect your app with firebase console 9
10. Steps to integrate MLKit library 2. Add MLKit dependencies to your Gradle file 10
11. Steps to integrate MLKit library 3. Convert input image to bitmap format and feed to FirebaseVisionDetector 11
12. Steps to integrate MLKit library 4. Parse and parse the results from the detection 12
13. Implementation - Summary 1. Connect your app with firebase 2. Add Gradle MLKit dependencies 3. Run your image through the recognizer for processing 4. Retrieve and parse the processed results 13
14. Thank you! You can reach me @sangee_kp on Twitter 14