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
Tensorflow for Android Developers
Search
Joe Birch
March 26, 2018
Technology
3
250
Tensorflow for Android Developers
Joe Birch
March 26, 2018
Tweet
Share
More Decks by Joe Birch
See All by Joe Birch
Learning to play guitar with Actions on Google
hitherejoe
1
96
Making Change as an Ally
hitherejoe
1
460
Learning to play the guitar with Actions on Google
hitherejoe
0
130
For Optimists, our UI is pretty Pessimistic
hitherejoe
4
2.8k
Android Things: Building for the IoT
hitherejoe
2
170
Getting Clean, Keeping Lean
hitherejoe
10
660
Android TV: Building Apps with Google’s Leanback Library
hitherejoe
1
1k
Building Beautiful Apps with the Design Support Library
hitherejoe
3
240
Other Decks in Technology
See All in Technology
AIのAIによるAIのための出力評価と改善
chocoyama
0
510
白金鉱業Meetup_Vol.19_PoCはデモで語れ!顧客の本音とインサイトを引き出すソリューション構築
brainpadpr
2
490
原則から考える保守しやすいComposable関数設計
moriatsushi
3
500
2年でここまで成長!AWSで育てたAI Slack botの軌跡
iwamot
PRO
2
140
Snowflake Summit 2025 データエンジニアリング関連新機能紹介 / Snowflake Summit 2025 What's New about Data Engineering
tiltmax3
0
230
Prox Industries株式会社 会社紹介資料
proxindustries
0
210
CSS、JSをHTMLテンプレートにまとめるフロントエンド戦略
d120145
0
220
より良いプロダクトの開発を目指して - 情報を中心としたプロダクト開発 #phpcon #phpcon2025
bengo4com
1
360
Navigation3でViewModelにデータを渡す方法
mikanichinose
0
210
IAMのマニアックな話 2025を執筆して、 見えてきたAWSアカウント管理の現在
nrinetcom
PRO
4
660
Welcome to the LLM Club
koic
0
130
データプラットフォーム技術におけるメダリオンアーキテクチャという考え方/DataPlatformWithMedallionArchitecture
smdmts
5
570
Featured
See All Featured
Into the Great Unknown - MozCon
thekraken
39
1.9k
Being A Developer After 40
akosma
90
590k
Bash Introduction
62gerente
614
210k
Fashionably flexible responsive web design (full day workshop)
malarkey
407
66k
It's Worth the Effort
3n
184
28k
The Myth of the Modular Monolith - Day 2 Keynote - Rails World 2024
eileencodes
26
2.8k
The Language of Interfaces
destraynor
158
25k
The MySQL Ecosystem @ GitHub 2015
samlambert
251
13k
The Cost Of JavaScript in 2023
addyosmani
51
8.4k
Large-scale JavaScript Application Architecture
addyosmani
512
110k
What’s in a name? Adding method to the madness
productmarketing
PRO
22
3.5k
Design and Strategy: How to Deal with People Who Don’t "Get" Design
morganepeng
130
19k
Transcript
TENSORFLOW FOR ANDROID DEVELOPERS JOE BIRCH - @HITHEREJOE - ANDROID
LEAD @BUFFER - GDE @ANDROID
MACHINE LEARNING 101 Get data Clean, prep & manipulate data
Train Model Test data Improve
MACHINE LEARNING 101 Unsupervised Learning Supervised Learning Clustering Classification Regression
MACHINE LEARNING 101 Unsupervised Learning Supervised Learning Clustering Classification Regression
MACHINE LEARNING 101 Unsupervised Learning Supervised Learning Clustering Classification Regression
MACHINE LEARNING 101 Unsupervised Learning Supervised Learning Clustering Classification Regression
MACHINE LEARNING 101 Unsupervised Learning Supervised Learning Clustering Classification Regression
MACHINE LEARNING 101 Unsupervised Learning Supervised Learning Clustering Classification Regression
MACHINE LEARNING AND MOBILE
MACHINE LEARNING AND MOBILE
TENSORFLOW
TENSORFLOW
COMPUTATION GRAPHS C D F A B E
NEURAL NETWORKS Some Image Result
NEURAL NETWORKS Some Image Result
NEURAL NETWORKS Some Image Result
NEURAL NETWORKS Some Image Result Pre-trained model
BUILDING OUR OWN IMAGE CLASSIFIER USING A MOBILE NET
TENSORBOARD
TENSORBOARD
TENSORBOARD tensorboard --logdir tf_files/training_summaries &
TRAINING DATA
RETRAINING THE MODEL
RETRAINING THE MODEL
RETRAINING THE MODEL
RETRAINING THE MODEL
RETRAINING THE MODEL
RETRAINING THE MODEL
RETRAINING THE MODEL
RETRAINING THE MODEL
RETRAINING THE MODEL
RETRAINING THE MODEL
RETRAINING THE MODEL
HOW TRAINING WORKS?
HOW TRAINING WORKS?
HOW TRAINING WORKS?
HOW TRAINING WORKS?
HOW TRAINING WORKS?
HOW TRAINING WORKS?
MODEL ACCURACY
MODEL ACCURACY
OPTIMISING THE MODEL Model Operation Model Operation Model Operation Model
Operation Load graph Don’t load graph Operation supported?
OPTIMISING THE MODEL
OPTIMISING THE MODEL
OPTIMISING THE MODEL
OPTIMISING THE MODEL
OPTIMISING THE MODEL
QUANTISATION
QUANTISATION
QUANTISATION
QUANTISATION
OPTIMISING THE MODEL
ADDING THIS TO AN APP Add Dependancy Create TF Reference
Feed data Run inference Fetch result Handle confidence
ADDING THIS TO AN APP
ADDING THIS TO AN APP // convert to 3d array
(width / height / color)
ADDING THIS TO AN APP
ADDING THIS TO AN APP
ADDING THIS TO AN APP Shape of our input
ADDING THIS TO AN APP
ADDING THIS TO AN APP
ADDING THIS TO AN APP
ADDING THIS TO AN APP https://github.com/tensorflow/tensorflow
CONCLUSION