Google ML-Kit - Understanding on-device machine learning
- 1. Understanding on-device Machine machine learning ML-Kit by Google Vishrut Goyani Android developer vishrutgoyani9 vishrut.goyani3
- 3. Tell me more, I'm listening • ML Kit is a mobile SDK that brings Google's on-device machine learning expertise to Android and iOS apps. • Use powerful yet easy to use Vision and Natural Language APIs to solve common challenges in your apps or create brand-new user experiences. • ML Kit's APIs all run on-device, allowing for real-time use cases where you want to process a live camera stream for example. This also means that the functionality is available offline.
- 4. Offerings • Text recognition • Face detection • Face mesh detection • Pose detection • Selfie segmentation • Subject segmentation • Document scanner • Barcode scanning • Image labelling • Object detection • Digital ink recognition • Language identification • Translation • Smart reply • Entity extraction Vision Natural language
- 5. Wait, there is one more!
- 6. Firebase ML
- 7. Firebase ML - Is it same?
- 8. What is it? • Firebase Machine Learning is a mobile SDK that brings Google's machine learning expertise to Android and Apple apps in a powerful yet easy-to-use package. • Firebase ML provides the ability to deploy custom models to your users' devices by uploading them to our servers. • It allows you to keep your app's initial install size small, and you can swap the ML model without having to republish your app.
- 9. Offerings • Text recognition • Image labeling • Object detection and tracking • Face detection and contour tracing • Barcode scanning • Language identification • Translation • Smart reply Vision Natural language
- 10. Well, Here's the difference ML-Kit • On-device machine learning for common tasks • Faster for simpler tasks • Excellent offline performance • Simpler to set up and use • Need to update the app when updating the mode Firebase ML • On-device and cloud-based machine learning • Higher accuracy for complex tasks • Limited offline performance • More complex for cloud-based models • Update your models on cloud without requiring the app update
- 12. Image labelling using ML-Kit
- 13. Let’s Cook! Ingredients : • Android Studio • Java/Kotlin • Knowledge of Asynchronous programming like Threads or Coroutines • Android device or emulator (Obviously the device will be better)
- 14. First things first - Dependencies!
- 17. Base model capabilities :-
- 18. Custom model implementation :- Download custom model Implementation guide
- 19. Dependencies:-
- 20. In your app level gradle file To ensure that gradle doesn’t compress the model file when building the app
- 21. Create Image object :-
- 24. To load the downloaded custom model
- 25. Make an instance of image labeler For base model: For custom model:
- 26. Run the image labeler :-
- 27. Get information about labeled objects Now, you can use this information to group images by label names
- 28. Finally, We have done it…!
- 29. Wait, You got 10000+ photos? I’m screwed.
- 30. Time to go asynchronous 🙌🏻
- 32. The difference • Execution of Threads is heavyweight • It’s complex and can lead to unexpected behavior upon cancellation • Best suitable for simple & long running tasks • OS managed • Basic tasks requires fine-grained control Threads Coroutines • Coroutines are lightweight • Structured and graceful cancellation • Can be used for complex tasks with efficiency • Dispatcher managed • Handle complex asynchronous tasks efficiently
- 33. Hence proved, Coroutine is the way to go 🙌🏻 Implementation : In your gradle file,
- 34. Basics : launch, async and runBlocking To work with coroutines, Kotlin provides three basic building blocks: launch, async, and runBlocking. 1. Launch Launch is used to fire and forget coroutine. It’s perfect for cases where you don’t need to compute any result.
- 35. 1. Async async is used when you need a result computed in a coroutine. It starts a new coroutine and returns a Deferred<T>, which is a non-blocking future that represents a promise to provide a result later.
- 36. 1. RunBlocking runBlocking is a bridge between non-coroutine world and coroutine world. It's a way to start top-level main coroutine.
- 37. Now, Let’s talk about dispatchers
- 38. Example :
- 39. Want to know more? Scan this code and go ahead 👍🏼
- 40. Time to apply it in our code
- 41. Replacing linear execution in coroutines
- 44. Now you can say, Yesss! It’s done 🤩
- 46. Thank you! I’m here vishrutgoyani9 vishrut_goyani9 vishrut.goyani3 vishrut.goyani9