Watson Visual Recognition with Core ML

Performing visual recognition on device can be a hassle if you're not a deep learning expert. Let me show you how to do it simply with Watson Visual Recognition with Core ML!

Transcript

1. Watson Visual Recognition with Core ML
   DI-2 Developer Advocate at IBM Yacine Rezgui (@yrezgui)
2. Average RTT on mobile is 300 ms compared to 50 ms on desktop Source: radware.com
3. Introducing a 500 ms network delay resulted in increases of up to 26% in user frustration
4. What is Core ML? Core ML is a machine learning framework that enable can be used in the Apple ecosystem. It’s sits on top of Accelerate, BNNS, and Metal Performance Shaders, API collections that enhance GPU and CPU mathematical computations.
5. What is Core ML? Core ML is optimized for on-device performance, which minimizes memory footprint and power consumption
6. Why classifying on the edge? Responsiveness Privacy Smart System I want to find my images easily I don’t want to wait ages to interact with an application Are my data reused?
7. Core ML integration with Watson Visual Recognition Watson Studio Apple Device User Developer Custom classifier 1- Upload images by classes 2- Generate model 3- Download model 5- Add more images to each class 4- Select image to be classified 6- Restart the app 7- Check if there is a new model 8- Download new model
8. Why choose Watson Visual Recognition + Core ML? • Open source mobile SDK • Online training tool easy to use • Compatible with all Apple devices • Classification with online/offline mode • No deep learning knowledges required • REST API with client libraries for NodeJS, Python & Java • Fast classification with average processing time < 100 ms • Local model downloaded automatically in the background
9. Demo time bit.ly/2FQYiKq
10. Demo time
11. Demo time
12. Training images
13. Demo time
14. Demo time
15. Demo time
16. Negative images
17. Demo time
18. Demo time
19. Demo time
20. Demo time
21. Demo time
22. Thank you! twitter.com/yrezgui