Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Who's Afraid Of Machine Learning? & first steps...
Search
Britt Barak
April 23, 2018
Technology
5
900
Who's Afraid Of Machine Learning? & first steps with TensorFlow
Chicago Roboto & Android Makers 2018
Britt Barak
April 23, 2018
Tweet
Share
More Decks by Britt Barak
See All by Britt Barak
[Vonage] Introducing Conversations
brittbarak
1
120
Kids, Play Nice! Kotlin-Java Interop In Mind
brittbarak
2
440
Sharing is Caring- Getting Started with Kotlin Multiplatform
brittbarak
2
2k
Between JOMO and FOMO: You are reshaping communication.
brittbarak
2
1.2k
Build Apps For The Ones You Love
brittbarak
1
120
What an ML-ful World! MLKit for Android dev.
brittbarak
0
130
Make your app dance with MotionLayout
brittbarak
8
1.4k
Who's afraid of ML? V2 : First steps with MlKit
brittbarak
1
450
Oh, the places you'll go! Cracking Navigation on Android
brittbarak
0
480
Other Decks in Technology
See All in Technology
人に寄り添うAIエージェントとアーキテクチャ #BetAIDay
layerx
PRO
8
2k
dipにおけるSRE変革の軌跡
dip_tech
PRO
1
230
専門分化が進む分業下でもユーザーが本当に欲しかったものを追求するプロダクトマネジメント/Focus on real user needs despite deep specialization and division of labor
moriyuya
1
1.1k
マルチモーダル基盤モデルに基づく動画と音の解析技術
lycorptech_jp
PRO
4
540
AI時代の経営、Bet AI Vision #BetAIDay
layerx
PRO
1
1.8k
Nx × AI によるモノレポ活用 〜コードジェネレーター編〜
puku0x
0
360
LIFF CLIとngrokを使ったLIFF/LINEミニアプリのお手軽実機確認
diggymo
0
230
いかにして命令の入れ替わりについて心配するのをやめ、メモリモデルを愛するようになったか(改)
nullpo_head
2
440
マルチプロダクト×マルチテナントを支えるモジュラモノリスを中心としたアソビューのアーキテクチャ
disc99
1
320
KubeCon + CloudNativeCon Japan 2025 Recap
donkomura
0
170
AI によるドキュメント処理を加速するためのOCR 結果の永続化と再利用戦略
tomoaki25
0
420
解消したはずが…技術と人間のエラーが交錯する恐怖体験
lamaglama39
0
190
Featured
See All Featured
Designing Experiences People Love
moore
142
24k
The Art of Delivering Value - GDevCon NA Keynote
reverentgeek
15
1.6k
Site-Speed That Sticks
csswizardry
10
750
Helping Users Find Their Own Way: Creating Modern Search Experiences
danielanewman
29
2.8k
Building Adaptive Systems
keathley
43
2.7k
Embracing the Ebb and Flow
colly
86
4.8k
Navigating Team Friction
lara
188
15k
[RailsConf 2023] Rails as a piece of cake
palkan
56
5.7k
Visualizing Your Data: Incorporating Mongo into Loggly Infrastructure
mongodb
47
9.6k
CSS Pre-Processors: Stylus, Less & Sass
bermonpainter
357
30k
Faster Mobile Websites
deanohume
308
31k
Templates, Plugins, & Blocks: Oh My! Creating the theme that thinks of everything
marktimemedia
31
2.5k
Transcript
Who’s afraid of Machine Learning? Britt Barak
Britt Barak Google Developer Expert - Android Women Techmakers Israel
Britt Barak @brittBarak
None
None
None
None
None
None
None
None
In a machine...
None
Strawberry Not Strawberry
Input Red Seeds pattern Top leaves 0.64 0.75 0.4
0.64 0.75 0.4 Input Red Seeds pattern Top leaves
0.64 0.75 0.4 Input Red Seeds pattern Top leaves
0.64 0.75 0.4 Input Red Seeds pattern Top leaves
0.64 0.75 0.4 Input 0.5 0.8 0.3 Red Seeds pattern
Top leaves
0.64 0.75 0.4 Input Red Seeds pattern Top leaves 0.5
0.8 0.3 0.5 * 0.64 + 0.8 * 0.75 + 0.3 * 0.4
0.64 0.75 0.4 Input Red Seeds pattern Top leaves 0.5
0.8 0.3 0.5 * 0.64 + 0.8 * 0.75 + 0.3 * 0.4 ___________ 1.04
0.64 0.75 0.4 Input Red Seeds pattern Top leaves 0.5
0.8 0.3 0.5 * 0.64 + 0.8 * 0.75 + 0.3 * 0.4 ___________ 1.04 + 0.7
0.64 0.75 0.4 1.74 0.5 * 0.64 + 0.8 *
0.75 + 0.3 * 0.4 ___________ 1.04 + 0.7 ___________ 1.74 Input Red Seeds pattern Top leaves 0.5 0.8 0.3
0.64 0.75 0.4 1.02 1.74 Input Red Seeds pattern Top
leaves 0.97
0.64 0.75 0.4 Input Red Seeds pattern Top leaves 1.02
1.74 0.97
0.64 0.75 0.4 Output Strawberry Not Strawberry Input Red Seeds
pattern Top leaves 1.02 1.74 0.97 0.87 0.13
0.64 0.75 0.4 0.87 0.13 Strawberry Not Strawberry Output Input
Red Seeds pattern Top leaves 1.02 1.74 0.97
None
0.7 0.03 0.01 3.72 0.89 1.92 Strawberry Not Strawberry Output
Input Red Seeds pattern Top leaves 0.2 0.8
0.7 0.03 0.01 3.72 0.89 1.92 Strawberry Not Strawberry Output
Input Red Seeds pattern Top leaves 0.2 0.8
0.7 0.03 0.01 3.72 0.89 1.92 0.2 0.8 Strawberry Not
Strawberry Output Input Red Seeds pattern Top leaves
0.5 * 0.64 + 0.8 * 0.75 + 0.3 *
0.4 ___________ 1.04 + 0.7 ___________ 1.74 Strawberry Not Not Strawberry Not Not Strawberry Not Not
Training TRAINING
0.64 0.75 0.4 1.02 1.74 0.97 0.89 0.11 Strawberry Not
Strawberry Output Input Red Seeds pattern Top leaves
Strawberry Not Strawberry Output Input Hidden Red Seeds pattern Top
leaves
None
Data science
We get a trained model !
TensorFlow - Open source - Widely used - Flexible for
scale: - 1 or more CPUs / GPUs - desktop, server, mobile device
Strawberry
Strawberry
Strawberry • Bandwidth • Performance • Latency • Network •
Security • Privacy • …
TensorFlow Mobile - Speech Recognition - Image Recognition - Object
Localization - Gesture Recognition - Translation - Text Classification - Voice Synthesis
Lightweight Fast Cross platform
MobileNet Inception-V3 SmartReply Models
None
Image Classifier classifier .classify(bitmap) label
1. Add Assets
None
labels.txt strawberry orange lemon fig pineapple banana jackfruit custard apple
pomegranate hay carbonara chocolate sauce dough meat loaf
2. Add TensorFlow Lite
repositories { maven { url 'https://google.bintray.com/tensorflow' } } dependencies
{ // ... implementation 'org.tensorflow:tensorflow-lite:+' } build.gradle
android { aaptOptions { noCompress "tflite" } } build.gradle
3. Create ImageClassifier.java
Image Classifier
ImageClassifier.java model = loadModelFile(); tflite = new Interpreter();
ImageClassifier.java model = loadModelFile(); tflite = new Interpreter(model);
MappedByteBuffer loadModelFile() { AssetFileDescriptor descriptor= getAssets().openFd(MODEL_PATH);
MappedByteBuffer loadModelFile() { AssetFileDescriptor descriptor= getAssets().openFd(MODEL_PATH); FileInputStream inputStream = new
FileInputStream(descriptor.getFileDescriptor()); FileChannel channel = inputStream.getChannel();
MappedByteBuffer loadModelFile() { AssetFileDescriptor descriptor= getAssets().openFd(MODEL_PATH); FileInputStream inputStream = new
FileInputStream(descriptor.getFileDescriptor()); FileChannel channel = inputStream.getChannel(); long start = descriptor.getStartOffset(); long length = descriptor.getDeclaredLength(); return channel.map(FileChannel.MapMode.READ_ONLY, start, length); }
Image Classifier [strawberry, apple, ... ] labels.txt
ImageClassifier.java model = loadModelFile(); tflite = new Interpreter(model); labelList =
loadLabelList();
labels.txt strawberry orange lemon fig pineapple banana jackfruit custard apple
pomegranate hay carbonara chocolate sauce dough meat loaf
List<String> loadLabelList() throws IOException { InputStreamReader inputStream = new InputStreamReader(getAssets().open(LABEL_PATH));
}
List<String> loadLabelList() throws IOException { InputStreamReader inputStream = new InputStreamReader(getAssets().open(LABEL_PATH));
BufferedReader reader = new BufferedReader(inputStream); }
List<String> loadLabelList() throws IOException { InputStreamReader inputStream = new InputStreamReader(getAssets().open(LABEL_PATH));
BufferedReader reader = new BufferedReader(inputStream); List<String> labelList = new ArrayList<>(); String line; while ((line = reader.readLine()) != null) { labelList.add(line); } }
List<String> loadLabelList() throws IOException { InputStreamReader inputStream = new InputStreamReader(getAssets().open(LABEL_PATH));
BufferedReader reader = new BufferedReader(inputStream); List<String> labelList = new ArrayList<>(); String line; while ((line = reader.readLine()) != null) { labelList.add(line); } reader.close(); return labelList; }
Image Classifier [ [0..6] , [ 0.1 ] , ...
] [strawberry, apple, ... ] probArray labels.txt
probArray = { [0.7], [0.3], [0], [0], } labelList =
{ strawberry, apple, pineapple, banana, } 0.3
ImageClassifier.java model = loadModelFile(); tflite = new Interpreter(model); labelList =
loadLabelList(); probArray = new float[1][labelList.size()];
Image Classifier [......] [ [0..6] , [ 0.1 ] ,
... ] [strawberry, apple, ... ] ByteBuffer probArray labels.txt
ImageClassifier.java model = loadModelFile(); tflite = new Interpreter(model); labelList =
loadLabelList(); probArray = new float[1][labelList.size()]; imgData = ByteBuffer.allocateDirect( DIM_IMG_SIZE_X * DIM_IMG_SIZE_Y * DIM_PIXEL_SIZE); imgData.order(ByteOrder.nativeOrder());
4. Run the model / classify
classifier .classify(bitmap) Image Classifier [......] [ [0..6] , [ 0.1
] , ... ] [strawberry, apple, ... ] ByteBuffer probArray labels.txt
ImageClassifier.java String classify(Bitmap bitmap) { convertBitmapToByteBuffer(imgData, bitmap);
}
void convertBitmapToByteBuffer(Bitmap bitmap) { //... bitmap.getPixels(intValues, 0, bitmap.getWidth(), 0, 0,bitmap.getWidth(),
bitmap.getHeight()); }
void convertBitmapToByteBuffer(Bitmap bitmap) { //... bitmap.getPixels(intValues, 0, bitmap.getWidth(), 0, 0,bitmap.getWidth(),
bitmap.getHeight()); int pixel = 0; for (int i = 0; i < DIM_IMG_SIZE_X; ++i) { for (int j = 0; j < DIM_IMG_SIZE_Y; ++j) { final int val = intValues[pixel++]; imgData.put((byte) ((val >> 16) & 0xFF)); imgData.put((byte) ((val >> 8) & 0xFF)); imgData.put((byte) (val & 0xFF)); } } }
void convertBitmapToByteBuffer(Bitmap bitmap) { //... bitmap.getPixels(intValues, 0, bitmap.getWidth(), 0, 0,bitmap.getWidth(),
bitmap.getHeight()); int pixel = 0; for (int i = 0; i < DIM_IMG_SIZE_X; ++i) { for (int j = 0; j < DIM_IMG_SIZE_Y; ++j) { final int val = intValues[pixel++]; imgData.put((byte) ((val >> 16) & 0xFF)); imgData.put((byte) ((val >> 8) & 0xFF)); imgData.put((byte) (val & 0xFF)); } } }
ImageClassifier.java String classify(Bitmap bitmap) { convertBitmapToByteBuffer(imgData, bitmap); tflite.run(imgData,
probArray); }
ImageClassifier.java String classify(Bitmap bitmap) { convertBitmapToByteBuffer(imgData, bitmap); tflite.run(imgData,
probArray); String textToShow = getTopLabels(); return textToShow; }
Strawberry - 0.87 Apple - 0.13 Tomato - 0.01
Machine Learning is a new world
Links - Tensorflow - https://www.tensorflow.org/ - Tensorflow lite - https://www.tensorflow.org/mobile/tflite/
- Codes labs - codelabs.developers.google.com/codelabs/tensorflow-for-poets-2-tflite/ - Google’s Machine Learning Crash Course - developers.google.com/machine-learning/crash-course/ - [Dr. Joe Dispenza]
Thank you! Keep in touch! Britt Barak @brittBarak