Slide 1

Slide 1 text

AI for Art & Design with TensorFlow Lite Margaret Maynard-Reid, Jan 31, 2021 @margaretmz

Slide 2

Slide 2 text

@margaretmz | #AIArt | #TensorFlow | #TFLite ML Engineer & Artist 2 https://margaretmz.art/

Slide 3

Slide 3 text

@margaretmz | #AIArt | #TensorFlow | #TFLite Topics ● Computer vision tasks ● Generative art & design ● Datasets & resources AI for Art & Design TensorFlow Lite Android Samples & tutorials ● Sample apps ● Awesome TFLite ● E2E TFLite Tutorials ● What is TFLite? ● TFLite on Android ● Tools & resources 3

Slide 4

Slide 4 text

@margaretmz | #AIArt | #TensorFlow | #TFLite Art & Design ● Photography ● Sculpture ● Drawing ● Painting ● Fashion design ● Product design ● UI/UX design ● ... 4

Slide 5

Slide 5 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 resolution - Image-to-image 5

Slide 6

Slide 6 text

Generative Art & Design AI for Art & Design with TFLite 6

Slide 7

Slide 7 text

@margaretmz | #AIArt | #TensorFlow | #TFLite Deep Dream 2015 7

Slide 8

Slide 8 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 8

Slide 9

Slide 9 text

@margaretmz | #AIArt | #TensorFlow | #TFLite Generative Adversarial Networks 2014 What are GANs? Generative Adversarial Networks (GANs) have at least two network models which compete against each other ... 9

Slide 10

Slide 10 text

@margaretmz | #AIArt | #TensorFlow | #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… 10

Slide 11

Slide 11 text

@margaretmz | #AIArt | #TensorFlow | #TFLite ● Training datasets must be paired ● A type of conditional GAN Link to TensorFlow tutorial Pix2Pix 2016 11

Slide 12

Slide 12 text

@margaretmz | #AIArt | #TensorFlow | #TFLite CycleGAN 2017 ● Training datasets are unpaired ● Horses <--> zebras ● Summer <--> winter ● Photos to paintings (Monet, Van Gogh etc.) Link to TensorFlow tutorial 12

Slide 13

Slide 13 text

@margaretmz | #AIArt | #TensorFlow | #TFLite BigGAN 2018 Large scale GAN that generates high fidelity class-conditional images. Link to paper | TF Hub biggan512, biggan-256, biggan-128 13

Slide 14

Slide 14 text

@margaretmz | #AIArt | #TensorFlow | #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 14

Slide 15

Slide 15 text

@margaretmz | #AIArt | #TensorFlow | #TFLite Sketch2Code Microsoft: Turn your whiteboard sketches to working code in seconds with Sketch2Code 15

Slide 16

Slide 16 text

@margaretmz | #AIArt | #TensorFlow | #TFLite Fashion++ Facebook AI: “Fashion++: Minimal Edits for Outfit Improvement” (news | paper) 16

Slide 17

Slide 17 text

@margaretmz | #AIArt | #TensorFlow | #TFLite 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 UI/UX Design ● Rico mobile app datasets ● Kaggle ○ Common Mobile/Web App Icons ○ Icons-50 17 https://github.com/margaretmz/awesome-ai-art-design

Slide 18

Slide 18 text

@margaretmz | #AIArt | #TensorFlow | #TFLite Magenta https://magenta.tensorflow.org/ | https://github.com/magenta 18

Slide 19

Slide 19 text

https://ami.withgoogle.com/ 19

Slide 20

Slide 20 text

@margaretmz | #AIArt | #TensorFlow | #TFLite AI & the Future of UX https://uxdesign.cc/ai/home 20

Slide 21

Slide 21 text

@margaretmz | #AIArt | #TensorFlow | #TFLite AI for Art & Design AI will assist and inspire artists and designers Everyone can now make art and design! 21

Slide 22

Slide 22 text

TensorFlow Lite AI for Art & Design with TFLite 22

Slide 23

Slide 23 text

@margaretmz | #AIArt | #TensorFlow | #TFLite What is an ML model? • A complex mathematical function • A representation of learnings from training data 23

Slide 24

Slide 24 text

@margaretmz | #AIArt | #TensorFlow | #TFLite Model training, conversion & inference 24 Model ● tf.Keras (TensorFlow) ● Python libraries: Numpy, Matplotlib etc SavedModel or Keras model Serving ● Cloud ● Web ● Mobile ● IoT ● Micro controllers ● Edge TPU Training Inference Data

Slide 25

Slide 25 text

@margaretmz | #AIArt | #TensorFlow | #TFLite Why on device? It’s all about that camera! 25

Slide 26

Slide 26 text

@margaretmz | #AIArt | #TensorFlow | #TFLite What is TensorFlow Lite? A framework with a set of tools for deploying ML models to mobile & embedded: ● A converter ● An interpreter ● Ops / kernels ● Interface to hardware accelerations: ○ NN API ○ Edge TPU ● A bunch of other tools: quantization, pruning, benchmarking... 26

Slide 27

Slide 27 text

@margaretmz | #AIArt | #TensorFlow | #TFLite 27

Slide 28

Slide 28 text

@margaretmz | #AIArt | #TensorFlow | #TFLite TensorFlow Lite Support 0.1.0 ● Task API ● Support Library ● Metadata Library ● Codegen ● Custom Ops ● Model Maker ● ML Model Binding 28

Slide 29

Slide 29 text

@margaretmz | #AIArt | #TensorFlow | #TFLite tf.Keras to TFLite to Android 29

Slide 30

Slide 30 text

@margaretmz | #AIArt | #TensorFlow | #TFLite Android app with TensorFlow Lite ● 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: canvas, gallery or camera ○ Load model & run inference ○ Display results in UI 30

Slide 31

Slide 31 text

@margaretmz | #AIArt | #TensorFlow | #TFLite TensorFlow Lite tools 1. Train a model ● Teacheable machine ● Model Maker ( w/ transfer learning) ● ML Kit AutoML Vision ● TensorFlow Hub 31 2. Convert ● Converter ● Optimization toolkit ● Benchmarking tool 3. Deploy ● ML Model Binding ● Task API

Slide 32

Slide 32 text

https://pair.withgoogle.com/ 32

Slide 33

Slide 33 text

Android Apps and Tutorials AI for Art & Design with TFLite 33

Slide 34

Slide 34 text

@margaretmz | #AIArt | #TensorFlow | #TFLite ML Kit Digital Ink Recognition Recognize handwritten text and drawings ● Text models: 300+ languages ● Non-text models ○ Autodraw ○ Emoji ○ Shapes Blog: Digital Ink Recognition in ML Kit Sample app: Overview | GitHub 34 Example: Draw & recognize emojis

Slide 35

Slide 35 text

@margaretmz | #AIArt | #TensorFlow | #TFLite ESRGAN on Android with TFLite ● Enhanced Super Resolution GAN ● Pre-trained ESRGAN model from TF Hub ● Convert to TFLite model ○ Dynamic range quantization ○ Input image 50x50 ● TensorFLow C API, GPU library ● Android NDK 35 https://blog.tensorflow.org/2020/12/how-to-generate-super -resolution-images-using-tensorflow-lite-on-android.html

Slide 36

Slide 36 text

@margaretmz | #AIArt | #TensorFlow | #TFLite Part 2: ML Model Binding from Android Studio TensorFlow Lite Model Maker Part 1: TensorFlow Lite Model Maker ● Image/Text classification 36

Slide 37

Slide 37 text

@margaretmz | #AIArt | #TensorFlow | #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 37

Slide 38

Slide 38 text

@margaretmz | #AIArt | #TensorFlow | #TFLite 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 38

Slide 39

Slide 39 text

@margaretmz | #AIArt | #TensorFlow | #TFLite Cartoonizer ● Model: 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 39

Slide 40

Slide 40 text

@margaretmz | #AIArt | #TensorFlow | #TFLite Segmentation + Style Transfer Demo by Khanh LeViet TFLite inference in Python, Already implemented in Android (and iOS)... (project repo) 40

Slide 41

Slide 41 text

@margaretmz | #AIArt | #TensorFlow | #TFLite 2020 E2E TFLite Tutorials GitHub repo 41

Slide 42

Slide 42 text

@margaretmz | #AIArt | #TensorFlow | #TFLite 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 42

Slide 43

Slide 43 text

@margaretmz | #AIArt | #TensorFlow | #TFLite Follow me on Twitter, Medium or GitHub to learn more about deep learning, TensorFlow and on-device ML. @margaretmz @margaretmz margaretmz Thank you! 43