Slide 1

Slide 1 text

AI for Art & Design with Lite Margaret Maynard-Reid, Sept 4 2020 #TFUGIndiaSummit @margaretmz

Slide 2

Slide 2 text

#TFUGIndiaSummit | @margaretmz | #TFLite Topics ● Computer vision tasks ● Generative art & design ● Datasets 2 AI for Art & Design TensorFlow Lite Samples & tutorials ● Samples & tutorials ● Community ● Tools & resources ● What is TFLite? ● How to make apps

Slide 3

Slide 3 text

Computer Vision Tasks Complexity Image classification (single-label) Classify an image to a class Examples Painting style or artist Van Gogh Image classification (multi-label) Classify an image to multiple classes Examples Movie poster genre action, sci-fi Feature extraction Extracting latent features of an image with CNN models Examples Visual search find similar fashion Object detection Identify one or multiple objects within an image and their locations with bounding boxes. detect UI elements Segmentation Classify whether each pixel of the image belongs to a certain class segment UI elements Generative models (GANs) Two or more models trained simultaneously Examples - Generate new images - Super res - Image-to-image 3

Slide 4

Slide 4 text

Generative Art & Design AI for Art & Design with TFLite

Slide 5

Slide 5 text

#TFUGIndiaSummit | @margaretmz | #TFLite Deep Dream 2015 5

Slide 6

Slide 6 text

Use CNN neural networks Content image + style image = styled image Content Style Generated Neural Style Transfer: 50 Shades of Miaw (link) A few more samples of generated images… Style Transfer 2015

Slide 7

Slide 7 text

#TFUGIndiaSummit | @margaretmz | #TFLite Generative Adversarial Networks 2014 What are GANs? Generative Adversarial Networks (GANs) have at least two network models which compete against each other ... 7

Slide 8

Slide 8 text

#TFUGIndiaSummit | @margaretmz | #TFLite GANs variations GANs are fun but difficult to train! ● 2014 Ian Goodfellow’s original GAN’s paper ● 2015 DCGAN (Deep Convolutional GAN) (paper) ● 2014 cGAN (Conditional GAN) - Generate images conditioned on a class (paper) ● 2016 Pix2Pix – paired image to image translation (paper) ● 2017 CycleGAN – unpaired Image to image translation ● 2018 StyleGAN – Generate images with some control on image characteristics (paper) ● Many other GAN variations… 8

Slide 9

Slide 9 text

#TFUGIndiaSummit | @margaretmz | #TFLite ● Training datasets must be paired ● A type of conditional GAN Link to TensorFlow tutorial Pix2Pix 2016 9

Slide 10

Slide 10 text

#TFUGIndiaSummit | @margaretmz | #TFLite CycleGAN 2017 ● Training datasets are unpaired ● Horses <--> zebras ● Summer <--> winter ● Photos to paintings (Monet, Van Gogh etc.) Link to TensorFlow tutorial 10

Slide 11

Slide 11 text

#TFUGIndiaSummit | @margaretmz | #TFLite BigGAN 2018 Large scale GAN that generates high fidelity class-conditional images. Link to paper | TF Hub biggan512, biggan-256, biggan-128 11

Slide 12

Slide 12 text

#TFUGIndiaSummit | @margaretmz | #TFLite Control the styles of the generated images ● Pose ● Face shape ● Age ● Gender ● Hair color 4x4 --> 1024x1024 8GPUs trained for 1 week Current examples: faces, bedrooms and cats Link to video StyleGAN 2018 12

Slide 13

Slide 13 text

#TFUGIndiaSummit | @margaretmz | #TFLite Sketch2Code Microsoft: Turn your whiteboard sketches to working code in seconds with Sketch2Code 13

Slide 14

Slide 14 text

#TFUGIndiaSummit | @margaretmz | #TFLite Fashion++ Facebook AI: “Fashion++: Minimal Edits for Outfit Improvement” (news | paper) 14

Slide 15

Slide 15 text

Datasets Art ● The Art Institute of Chicago - 5000 high res images ● Best Artworks of All Time ● Art Images: Drawing/Painting/Sculptures/Engravings ● Overwatch Heroes Recognition ● Chinese Fine Art ● Museum of Modern Art Collection 15 UI/UX Design ● Rico mobile app datasets ● Kaggle ○ Common Mobile/Web App Icons ○ Icons-50

Slide 16

Slide 16 text

TensorFlow Lite AI for Art & Design with TFLite

Slide 17

Slide 17 text

#TFUGIndiaSummit | @margaretmz | #TFLite What is TensorFlow Lite? A framework with a set of tools for deploying ML to mobile & embedded: ● A converter ● An interpreter ● Ops / kernels ● Interface to hardware accelerations: ○ NN AIP ○ Edge TPu ● A bunch of other tools... 17

Slide 18

Slide 18 text

#TFUGIndiaSummit | @margaretmz | #TFLite Why on device? 18 It’s all about that camera!

Slide 19

Slide 19 text

#TFUGIndiaSummit | @margaretmz | #TFLite How to make a TensorFlow Lite app? ● ML model: ○ Train / find a model ○ Convert to .tflite format ○ Run inference in Python ○ Benchmark (optional) ○ Add model metadata ● Build Android app ○ UI navigation ○ Input image: gallery or camera ○ Run model inference ○ Display result in UI 19

Slide 20

Slide 20 text

#TFUGIndiaSummit | @margaretmz | #TFLite What is an ML model? • A complex mathematical function • A representation of learnings from training data 20

Slide 21

Slide 21 text

#TFUGIndiaSummit | @margaretmz | #TFLite How to get an ML model? ● Train your own ● Transfer learning ● Download from TensorFlow Hub ● Use ML Kit 21

Slide 22

Slide 22 text

https://pair.withgoogle.com/

Slide 23

Slide 23 text

Apps and Tutorials AI for Art & Design with TFLite

Slide 24

Slide 24 text

#TFUGIndiaSummit | @margaretmz | #TFLite ML Kit Digital Ink Recognition Draw & recognize emojis Blog post: https://developers.googleblog.com/2020/08/digital-ink- recognition-in-ml-kit.html Sample app: https://developers.google.com/ml-kit/vision/digital-ink-r ecognition 24

Slide 25

Slide 25 text

Part 2: ML Model Binding from Android Studio TensorFlow Lite Model Maker Part 1: TensorFlow Lite Model Maker ● Image/Text classification 25

Slide 26

Slide 26 text

#TFUGIndiaSummit | @margaretmz | #TFLite Style transfer Google Arts & Culture app: https://play.google.com/store/apps/details?id=com.google.android.apps.cultural Blog post: 4/2/2020 - Transform your photo in the style of an iconic artist 26

Slide 27

Slide 27 text

27 Selfie2Anime ● U-GAT-IT paper ● Image to image translation ● Written in TF1 code ● Unpaired selfie and anime images ● 50 & 100 epochs checkpoints ● Collaboration with Sayak & Khanh ● Links: project repo | blog

Slide 28

Slide 28 text

#TFUGIndiaSummit | @margaretmz | #TFLite Whitebox CartoonGAN ● Faster & smaller ● Not just for selfies ● Link to project repo ● Link to TF Hub models ● Collaboration with Sayak & Khanh ● Links: project repo | blog 28

Slide 29

Slide 29 text

#TFUGIndiaSummit | @margaretmz | #TFLite Segmentation + Style Transfer Demo by Khanh LeViet TFLite inference in Python, to be implemented on mobile... (project repo) 29

Slide 30

Slide 30 text

Community & resources AI for Art & Design with TFLite

Slide 31

Slide 31 text

#TFUGIndiaSummit | @margaretmz | #TFLite 2020 E2E TFLite Tutorials 31 Project repo

Slide 32

Slide 32 text

Awesome TensorFlow Lite A curated list of models, apps, tutorials and learning resources… Thanks to the ML GDE and Android community! Support of TensorFlow Lite team https://github.com/margaretmz/awesome-tflite 32

Slide 33

Slide 33 text

#TFUGIndiaSummit | @margaretmz | #TFLite Magenta https://magenta.tensorflow.org/ | https://github.com/magenta 33

Slide 34

Slide 34 text

https://ami.withgoogle.com/ 34

Slide 35

Slide 35 text

#TFUGIndiaSummit | @margaretmz | #TFLite AI & the Future of UX https://uxdesign.cc/ai/home 35

Slide 36

Slide 36 text

#TFUGIndiaSummit | @margaretmz | #TFLite AI for Art & Design AI will assist and inspire artists and designers Everyone can now make art and design! 36

Slide 37

Slide 37 text

#TFUGIndiaSummit | @margaretmz | #TFLite Follow me on Twitter, Medium or GitHub to learn more about deep learning, TensorFlow and on-device ML. @margaretmz @margaretmz margaretmz Thank you! 37