×
Copy
Open
Link
Embed
Share
Beginning
This slide
Copy link URL
Copy link URL
Copy iframe embed code
Copy iframe embed code
Copy javascript embed code
Copy javascript embed code
Share
Tweet
Share
Tweet
Slide 1
Slide 1 text
No content
Slide 2
Slide 2 text
How to AI in JS? Asim Hussain Developer Advocate Microsoft AI JavaScript Rocks
Slide 3
Slide 3 text
Asim Hussain @jawache codecraft.tv microsoft.com
Slide 4
Slide 4 text
https://aka.ms/jawache-cda @jawache
Slide 5
Slide 5 text
@jawache
Slide 6
Slide 6 text
Asim Web Development Machine Learning This is @EleanorHaproff's slide
Slide 7
Slide 7 text
No content
Slide 8
Slide 8 text
Opinion @jawache
Slide 9
Slide 9 text
TheMojifier™ @jawache
Slide 10
Slide 10 text
No content
Slide 11
Slide 11 text
@jawache
Slide 12
Slide 12 text
Calculate Emotion @jawache
Slide 13
Slide 13 text
https://towardsdatascience.com/facial-keypoints-detection-deep-learning-737547f73515
Slide 14
Slide 14 text
Use an Artificial Neural Network @jawache
Slide 15
Slide 15 text
Artificial Neural Networks Axon Dendrites Axons Body @jawache
Slide 16
Slide 16 text
Artificial Neural Networks 1 23 8.6 -0.5 2.1 Activation Function @jawache
Slide 17
Slide 17 text
Artificial Neural Networks 1 23 8.6 -0.5 2.1 x x activation(...) = -11.5 = 18.06 7.01 !-> !-> } @jawache
Slide 18
Slide 18 text
Output 0 0 1 Input Artificial Neural Networks @jawache
Slide 19
Slide 19 text
Output 0 -1 1 Input Artificial Neural Networks TanH @jawache
Slide 20
Slide 20 text
Output 0 0 1 Input Artificial Neural Networks Relu @jawache
Slide 21
Slide 21 text
Artificial Neural Networks 1.1 4.2 0.3 4 12 93 3 @jawache
Slide 22
Slide 22 text
Artificial Neural Networks 1.1 4.2 0.3 4 12 93 3 - 8 = -5 @jawache
Slide 23
Slide 23 text
Artificial Neural Networks 1.1 4.2 0.3 4 12 93 3 - 8 = -5 @jawache
Slide 24
Slide 24 text
Artificial Neural Networks 0.1 9.2 0.2 4 12 93 8 @jawache
Slide 25
Slide 25 text
@jawache https://azure.microsoft.com/services/cognitive-services/face/
Slide 26
Slide 26 text
https:!//.api.cognitive.microsoft.com/face/v1.0/detect { "url": "" } @jawache
Slide 27
Slide 27 text
@jawache
Slide 28
Slide 28 text
Summary @jawache
Slide 29
Slide 29 text
• Neural Networks are incredibly powerful • Conceptually, they are simple to understand @jawache Summary
Slide 30
Slide 30 text
TensorFlow, MobileNet & I'm fine @jawache
Slide 31
Slide 31 text
@jawache
Slide 32
Slide 32 text
@jawache
Slide 33
Slide 33 text
@jawache
Slide 34
Slide 34 text
TensorFlow.js @jawache
Slide 35
Slide 35 text
TensorFlow.js Train models Load pre-trained models @jawache
Slide 36
Slide 36 text
https://github.com/tensorflow/tfjs-models @jawache MobileNet
Slide 37
Slide 37 text
https://azure.microsoft.com/services/cognitive-services/computer-vision/ @jawache
Slide 38
Slide 38 text
https://codepen.io/sdras/full/jawPGa/ @jawache
Slide 39
Slide 39 text
@jawache https://twitter.com/ollee/status/930303340516216832
Slide 40
Slide 40 text
@jawache https://twitter.com/FrontendNE/status/930120267992616960
Slide 41
Slide 41 text
@jawache https://twitter.com/chrispiecom/status/930407801402347520
Slide 42
Slide 42 text
Summary @jawache
Slide 43
Slide 43 text
• TensorFlow.js doesn't have any dependancies • MobileNet is a simple way to analyse images • Azure Computer Vision API ❤ @jawache Summary
Slide 44
Slide 44 text
Image2Image @jawache
Slide 45
Slide 45 text
DEMO @jawache https://zaidalyafeai.github.io/pix2pix/cats.html
Slide 46
Slide 46 text
@jawache Generator Discriminator ✅ ❌
Slide 47
Slide 47 text
@jawache Generator Discriminator ✅ ❌
Slide 48
Slide 48 text
@jawache Generator Discriminator ✅ ✅
Slide 49
Slide 49 text
@jawache
Slide 50
Slide 50 text
@jawache
Slide 51
Slide 51 text
@jawache
Slide 52
Slide 52 text
@jawache https://github.com/NVIDIA/vid2vid
Slide 53
Slide 53 text
@jawache https://github.com/NVIDIA/vid2vid
Slide 54
Slide 54 text
@jawache https://github.com/NVIDIA/vid2vid
Slide 55
Slide 55 text
https://github.com/NVIDIA/vid2vid @jawache
Slide 56
Slide 56 text
@jawache https://github.com/hanzhanggit/StackGAN
Slide 57
Slide 57 text
Summary @jawache
Slide 58
Slide 58 text
• GANs learn to generate new images • They take a lot of compute to train • But the generator model can be run in the browser @jawache Summary
Slide 59
Slide 59 text
@jawache aka.ms/mojifier
Slide 60
Slide 60 text
@jawache themojifer.com
Slide 61
Slide 61 text
Tero Parviainen creative.ai Music and AI in the Browser with TensorFlow.js and Magenta.js
Slide 62
Slide 62 text
Thoughts @jawache
Slide 63
Slide 63 text
Asim Hussain @jawache codecraft.tv microsoft.com