AI and ML in mobile apps

I gave a talk on the basics of Artificial Intelligence and Machine Learning in Android Developers Meetup in Gurgaon, India.
In this session I explained the basics of AI/ML, how ML is different from AI and also give brief introduction to Deep learning. I explained how ML works and what are basic types of ML - Supervised Learning, unsupervised learning and Reinforcement learning. What are the applications and most recent prominent examples of ML. And then I moved on to introduce the major frameworks for developers in this field. I gave brief introduction to Google Cloud vision, OpenCV, Pytorch, AWS Rekognition and finally ML Kit from Google. I explained ML kit in detail and how developers can use it. I also gave a demo on ML kit.

Transcript

1. AI/ML and Mobile
2. Hello! I am Paramvir Singh I am Android developer and consultant. You can find me at: @paramvirsingh88 2
3. Agenda ◎ Understand AI and ML ◎ Different tools and frameworks ◎ Using ML in mobile apps - ML Kit 3
4. 1. Artificial Intelligence and Machine Learning 4
5. “ Artificial intelligence would be the ultimate version of Google. The ultimate search engine that would understand everything on the web. —Larry Page 5
6. AI A branch of computer science the objective of which is to make machine work as smart as human. Cognitive ability of machines. AI vs ML ML A computer program is fed a lot of data and learns from experiences. A subset of AI. 6
8. A new Tech? Large datasets, advancements in data science, powerful hardware 8
9. Types of Learning Supervised Output is fed into the system. Labeled data. Unsupervised Success and failure is not known to the system. No labeling. Pattern recognition. Reinforcement Once result is obtained, it is categorized as success or failure. Similar to unsupervised. 9
10. Understanding Machine Learning ◎ Convolution matrix ◎ Deep neural network, tensors ◎ Dynamic computation, computational graphs ◎ Automatic differentiation ◎ Relation Extraction, Coreference Resolution 10
11. Understanding Machine Learning ◎ Convolution matrix ◎ Deep neural network, tensors ◎ Dynamic computation, computational graphs ◎ Automatic differentiation ◎ Relation Extraction, Coreference Resolution 11
12. Instead we will go the easier route ◎ Data ◎ Model ◎ Training ◎ Learning ◎ Experience ◎ Decision 12
13. How ML models work Train Model Process Data Apply model 13
14. How it’s all spun together • Huge datasets required • Create and train models • Choose best model • Apply the model 14
15. ML models • 15
16. ML models • 16
17. Applications • Image processing • NLP and Speech recognition • Predictions ◦ Diagnostic ◦ Business and customers ◦ Natural events • Robotics/real time decisions 17
18. Some recent examples ▪ First AI art ▪ Cambridge Analytica case ▪ Cancer detection ▪ Self driving cars ▪ Google Translate and email smart reply 18
19. ML frameworks ▪ Cloud Vision ▪ ML kit ▪ TensorFlow ▪ Amazon Rekognition ▪ Facebook Pytorch ▪ OpenCV ▪ Azure cloud ML 19
20. ML Kit Easy Machine Learning 20
21. ML Kit Mobile focused Younger sibling of Google Cloud Vision. API Based No mode creation or model training required. Uses models provided and hosted by Google. Focussed on Vision but will expand in future. 21
22. What problem does it solve? ▪ Doing ML from scratch is doable but hard ▪ Huge data required ▪ Knowledge of math and ML science required ▪ Domain knowledge required 22
23. What ML Kit provides Model hosting in cloud Firebase API’s On device ML processing 23
24. ML Kit features Image Labeling Text Detection Face detection Barcode scanning Landmark detection Smart Reply* 24 *coming soon
25. Integration in Android Integrate Firebase and you are good to go. 25
26. Thanks! Any questions? You can find me at: ◎ @paramvirsingh88 ◎ paramvir.singh88@gmail.com 26