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

Deep Learning for Everyone (Rday Medellin)

Deep Learning for Everyone (Rday Medellin)

As a Data Scientist (or aspiring Data Scientist) we are overwhelmed by the amount of knowledge we need to have and acquire. Every day there is a new technique, a new framework, a new state of the art model. For the last few years, Deep Learning has become a hot topic and it is the main driver of many applications. But how can we start our Deep Learning journey? Which of the several deep learning frameworks should we use? Where can I find examples of code that work and that I can use without worrying about the license?

In this talk, I will show you how you can start with Deep Learning without any previous Deep Learning knowledge and how you can have a basic ready-to-use deep learning “service” running in less than five minutes.

Gabriela de Queiroz

November 08, 2019
Tweet

More Decks by Gabriela de Queiroz

Other Decks in Technology

Transcript

  1. Deep Learning for Everyone
    Gabriela de Queiroz
    Sr. Engineering & Data Science Manager, IBM
    Founder, R-Ladies & AI Inclusive
    @gdequeiroz | k-roz.com
    slides: bit.ly/rday-medellin

    View full-size slide

  2. Gabriela de Queiroz
    • Founder of R-Ladies
    • Founder of AI Inclusive (ai-inclusive.org)
    • Member of the R Foundation
    • Sr. Engineering & Data Science Manager, IBM
    Data Scientist + Developer Advocate + Open Source Developer +
    Manager + Community Builder + Mentor
    slides: bit.ly/rday-medellin

    View full-size slide

  3. slides: bit.ly/rday-medellin
    Worldwide organization that promotes
    diversity in the #rstats community
    via meetups and mentorship in a
    friendly and safe environment.

    View full-size slide

  4. slides: bit.ly/rday-medellin
    2012 - From Brazil to San Francisco

    View full-size slide

  5. 5
    San Francisco, CA
    October 2012

    View full-size slide

  6. slides: bit.ly/rday-medellin
    meetup.com/rladies-medellin
    facebook.com/MedellinRLadies
    twitter.com/RLadiesMedellin
    @gdequeiroz | www.k-roz.com

    View full-size slide

  7. Center for Open Source Data
    and AI Technologies
    (CODAIT)
    30+ open source developers!
    Watson West Building
    505 Howard St.
    San Francisco, California
    Improving Enterprise AI lifecycle in Open Source
    Gather
    Data
    Analyze
    Data
    Machine
    Learning
    Deep
    Learning
    Deploy
    Model
    Maintain
    Model
    Python
    Data Science
    Stack
    Fabric for
    Deep Learning
    (FfDL)
    Mleap +
    PFA
    Scikit-Learn
    Pandas
    Apache
    Spark
    Apache
    Spark
    Jupyter
    Model
    Asset
    eXchange
    Keras +
    Tensorflow
    CODAIT
    codait.org
    Gather
    Data
    Analyze
    Data
    Machine
    Learning
    Deep Learning
    Deploy
    Model
    Maintain
    Model
    Python
    Data Science
    Stack
    Fabric for Deep
    Learning
    (FfDL)
    PFA, PMML,
    ONNX
    Scikit-Learn
    Pandas
    Apache
    Spark
    Jupyter
    Model Asset
    eXchange
    (MAX)
    Tensorflow
    + PyTorch
    AIF360
    ART
    AIF360
    ART
    AIF360
    ART
    Apache
    Spark
    Data Asset
    eXchange
    (DAX)
    Build tools to make
    AI accessible to all
    @gdequeiroz | www.k-roz.com

    View full-size slide


  8. https://twitter.com/JeffDean/status/1135114657344237568?s=20

    View full-size slide

  9. > 4 million results!
    > 183 million results!

    View full-size slide

  10. slides: bit.ly/rday-medellin
    Help!

    View full-size slide

  11. Model Asset eXchange
    Place for developers/data scientists to find and use
    free and open source deep learning models
    ibm.biz/model-exchange
    @gdequeiroz | www.k-roz.com

    View full-size slide

  12. 30+ ready to use deep learning models

    View full-size slide

  13. Model Asset eXchange (MAX)
    • Wide variety of domains (text, audio, image, etc)
    • Multiple deep learning frameworks (TensorFlow, 

    PyTorch, Keras)
    • Trainable and Deployable versions
    ibm.biz/model-exchange

    View full-size slide

  14. What do I need to get started?

    View full-size slide

  15. https://www.docker.com

    View full-size slide

  16. Ways of accessing the models

    View full-size slide

  17. OBJECT DETECTOR
    Localize and identify multiple objects in a single image
    @gdequeiroz | www.k-roz.com

    View full-size slide

  18. ibm.biz/model-exchange

    View full-size slide

  19. Access the API via Swagger

    View full-size slide

  20. Access the API via Python
    Try yourself here:
    ibm.biz/max-notebook

    View full-size slide

  21. Access the API via R
    slides: bit.ly/rday-medellin

    View full-size slide

  22. Access the API via Web App
    Try yourself here:
    ibm.biz/object-detector-webapp

    View full-size slide

  23. Access the API via Node-RED flow

    View full-size slide

  24. Access the API via CodePen

    View full-size slide

  25. All this in a
    standardized way

    View full-size slide

  26. Find* a state-of-art open source deep
    learning model specific to domain
    Validate license terms
    Perform model health check &
    code clean up
    Wrap models in MAX framework and
    provide REST API
    Publish the deployable model as
    Docker images on Docker Hub
    Use the MAX training framework to
    create an image for custom model
    training
    Review and Continuous Integration
    * or build from scratch
    BEHIND THE SCENES

    View full-size slide

  27. And if you are
    feeling
    adventurous…

    View full-size slide

  28. You can train your model using your own data

    View full-size slide

  29. www.k-roz.com
    @gdequeiroz | www.k-roz.com

    View full-size slide

  30. How do I get started?

    View full-size slide

  31. @gdequeiroz | www.k-roz.com
    ibm.biz/max-tutorial

    View full-size slide

  32. Code Patterns
    How to easily consume MAX models
    ibm.biz/max-code-patterns

    View full-size slide

  33. Would you like to contribute?
    Check our central repository containing all details about
    contribution.
    ibm.biz/max-central-repo

    View full-size slide

  34. Ideas for contribution:
    1) New model using the Model Asset Exchange Skeleon
    ibm.biz/max-central-repo

    View full-size slide

  35. Ideas for contribution:
    2) Demo notebooks in Python
    ibm.biz/max-central-repo

    View full-size slide

  36. Ideas for contribution:
    3) Demo notebooks (.Rmd for example) in R
    4) Shiny Apps
    ibm.biz/max-central-repo
    ibm.biz/object-detector-webapp

    View full-size slide

  37. Photo by Peter Adams -
    Check this out: http://www.facesofopensource.com/
    I'm the first one to be representing #rstats #rladies
    Thank you!
    K-ROZ .COM
    @GDEQUEIROZ
    ai-inclusive.org
    ibm.biz/model-exchange

    View full-size slide