Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Tensorflow for Android Developers

Tensorflow for Android Developers

1cf799036b5d9439e9ed823c9b0c15cb?s=128

Joe Birch

March 26, 2018
Tweet

Transcript

  1. TENSORFLOW FOR ANDROID DEVELOPERS JOE BIRCH - @HITHEREJOE - ANDROID

    LEAD @BUFFER - GDE @ANDROID
  2. MACHINE LEARNING 101 Get data Clean, prep & manipulate data

    Train Model Test data Improve
  3. MACHINE LEARNING 101 Unsupervised Learning Supervised Learning Clustering Classification Regression

  4. MACHINE LEARNING 101 Unsupervised Learning Supervised Learning Clustering Classification Regression

  5. MACHINE LEARNING 101 Unsupervised Learning Supervised Learning Clustering Classification Regression

  6. MACHINE LEARNING 101 Unsupervised Learning Supervised Learning Clustering Classification Regression

  7. MACHINE LEARNING 101 Unsupervised Learning Supervised Learning Clustering Classification Regression

  8. MACHINE LEARNING 101 Unsupervised Learning Supervised Learning Clustering Classification Regression

  9. MACHINE LEARNING AND MOBILE

  10. MACHINE LEARNING AND MOBILE

  11. TENSORFLOW

  12. TENSORFLOW

  13. COMPUTATION GRAPHS C D F A B E

  14. NEURAL NETWORKS Some Image Result

  15. NEURAL NETWORKS Some Image Result

  16. NEURAL NETWORKS Some Image Result

  17. NEURAL NETWORKS Some Image Result Pre-trained model

  18. BUILDING OUR OWN IMAGE CLASSIFIER USING A MOBILE NET

  19. TENSORBOARD

  20. TENSORBOARD

  21. TENSORBOARD tensorboard --logdir tf_files/training_summaries &

  22. TRAINING DATA

  23. RETRAINING THE MODEL

  24. RETRAINING THE MODEL

  25. RETRAINING THE MODEL

  26. RETRAINING THE MODEL

  27. RETRAINING THE MODEL

  28. RETRAINING THE MODEL

  29. RETRAINING THE MODEL

  30. RETRAINING THE MODEL

  31. RETRAINING THE MODEL

  32. RETRAINING THE MODEL

  33. RETRAINING THE MODEL

  34. HOW TRAINING WORKS?

  35. HOW TRAINING WORKS?

  36. HOW TRAINING WORKS?

  37. HOW TRAINING WORKS?

  38. HOW TRAINING WORKS?

  39. HOW TRAINING WORKS?

  40. MODEL ACCURACY

  41. MODEL ACCURACY

  42. OPTIMISING THE MODEL Model Operation Model Operation Model Operation Model

    Operation Load graph Don’t load graph Operation supported?
  43. OPTIMISING THE MODEL

  44. OPTIMISING THE MODEL

  45. OPTIMISING THE MODEL

  46. OPTIMISING THE MODEL

  47. OPTIMISING THE MODEL

  48. QUANTISATION

  49. QUANTISATION

  50. QUANTISATION

  51. QUANTISATION

  52. OPTIMISING THE MODEL

  53. ADDING THIS TO AN APP Add Dependancy Create TF Reference

    Feed data Run inference Fetch result Handle confidence
  54. ADDING THIS TO AN APP

  55. ADDING THIS TO AN APP // convert to 3d array

    (width / height / color)
  56. ADDING THIS TO AN APP

  57. ADDING THIS TO AN APP

  58. ADDING THIS TO AN APP Shape of our input

  59. ADDING THIS TO AN APP

  60. ADDING THIS TO AN APP

  61. ADDING THIS TO AN APP

  62. ADDING THIS TO AN APP https://github.com/tensorflow/tensorflow

  63. CONCLUSION