+
=
2000 - 2010 2010 - 2017 2017 - 2019 … 2019 …
.Net Platform
ASP.NET,
C#, WCF, WPF,
Silverlight
ObjC, Swift, MacOS
AWS, NodeJs, Python
React, NodeJs, Swift
MacOS, Kotlin,
AWS, Flutter
ML, AR, VR
Human Learning
Human Transfer Learning
Agenda
- Machine Learning Basics
- TensorFlow Training Environment
- TensorFlow + Keras Basics
- Keras API Basics
- TensorFlowLite Basics
- Flutter App Architecture
- Demo
Computer Science
Artificial Intelligence
Machine Learning
Deep Learning
CNN
Computer Science is the
study of computers and
computational systems.
AI is the simulation of
human intelligence processe
s by computer systems. 
Machine learning is a
method of data analysis that
automates analytical model.
Deep learning is a
Machine Learning technique
that teaches computers to
learn by example.
Machine learning (ML) is the scientific study of algorithms and statistical
models that computer systems use to perform a specific task without using
explicit instructions, relying on patterns and inference instead
Supervised Unsupervised Semi-Supervised
?
Input data
labeled
? ?
Input data
not labeled
? ?
Input data
mixed
Machine Learning algorithms
organized by learning style.
Machine Learning algorithms
organized by type.
Source: https://machinelearningmastery.com
Cognitive computing 

and Machine Learning
Image source: http://www.cognub.com/index.php/cognitive-platform/
Image Classification
Image source: Fei-Fei Li & Andrej Karpathy & Justin Johnson
Image Classification
Image source: Fei-Fei Li & Andrej Karpathy & Justin Johnson
Image Classification
Image source: Fei-Fei Li & Andrej Karpathy & Justin Johnson
Spatial Localization and Detection
Image source: Fei-Fei Li & Andrej Karpathy & Justin Johnson
Spatial Localization and Detection
Image source: Fei-Fei Li & Andrej Karpathy & Justin Johnson
Spatial Localization and Detection
Image source: Fei-Fei Li & Andrej Karpathy & Justin Johnson
Machine Learning Cloud Platforms
Deep Learning Frameworks
The core open source library to help you develop and
train ML models
Keras is a high-level neural networks API, written in Python and
capable of running on top of TensorFlow, CNTK, or Theano.
+ =
tf.keras
TensorFlow’s High Level API
Transfer Learning
Pre-trained
CNN
1000 Classifications
Pre-trained
CNN
New Training Data
+
New Classifications
TensorFlow Environment Setup
$ docker run -it --rm -v $(realpath ~/
notebooks):/tf/notebooks -p 8888:8888
tensorflow/tensorflow:latest-py3-jupyter
TF Keras

Simple Sequential Model
Sequential
Model
Save
Model
Convert
Model
TFLite
Model
TF Keras

Retraining Model
Pre-trained
Model New Model
Mobilenet
Optimize
Model
Save
Model
Convert
Model
TFLite
Model
TensorFlow Lite is an open source deep learning
framework for on-device inference.
How it works
Create/Find
Model
Convert
Model
Deploy Optimize
Image source: https://www.tensorflow.org/lite
TensorFlow Lite File
Image source: https://www.tensorflow.org/lite
FlatBuffers is an efficient cross platform serialization
library for C++, C#, C, Go, Java, JavaScript, Lobster, Lua,
TypeScript, PHP, Python, and Rust. It was originally
created at Google for game development and other
performance-critical applications.
FlatBuffers
TensorFlow Lite File
Image source: https://www.tensorflow.org/lite
Flutter is Google’s UI toolkit for building beautiful, natively compiled
applications for mobile, web, and desktop from a single codebase.
TensorFlow Mobile Workflow
Platform
Channel
tflite package
Interpretor
Flutter App
Operators
Flutter Architecture
Flutter Architecture
App Architecture
Flutter
Controller
iOS
Android
Main.dart
Router
Main
Module
Object
Detection
Module
Shared
Module
Home
Camera
App Screens
App Dependencies
pubspec.yaml
App Assets
pubspec.yaml
App Bootstrap
Main.dart
App Screens
App Screens
TensorFlow Environment Setup
DEMO TIME
Contact

Flutter + tensor flow lite = awesome sauce