TensorFlow.js — Musical Unit Signature Instant reCognition

Machine Learning - OMR

1. To Do

  1. Add favico
  2. Add fontawesome
  3. Auto open accordenon on step in URL
  4. Add Upload (Step 1)
  5. Add "Categories / classes" to upload
  6. Add Upload Partials (Step 3)
  7. Scale Upload Partials (Step 3)
  8. Add Show images / data.js ?
  9. Add training responses
  10. Add Crowd sourcing View
  11. Add Crowd sourcing Process
  12. Improve Crowd sourcing Layout
  13. Improve loading speed
  14. Fix alert function
  15. Local CDN
  16. Auto number accordion
  17. Partials > append original data to json file
  18. Show preview > outline already sent Partials
  19. Accordeon vervangen: https://mdbootstrap.com/snippets/jquery/mdbootstrap/670810
  20. Project opschonen
    • Cookie class
    • CSS opdelen: drawer / file upload
    • JS opdelen
  21. Musical Unit Signature Instant reCognition - MNIST-database class
    • show labels => json ? To be used in html view / buttons
    • create sprite
      • Fix display bug
      • Grayscaling ?
      • Centering image ?
    • create binary labels
    • create sprite
      • Fix display bug
      • Grayscaling ?
      • Centering image ?
    • Status
      • image met label
      • aantal responses per file
  22. Set Cookie for WebGL according to test
    http://tompiedom.net/play-a-long/ml.html
  23. Show / filter crowd data
    https://www.js-tutorials.com/jquery-tutorials/reading-csv-file-using-jquery/
  24. CrowdSourcing pagina:
    Div dynamisch opbouwen, betere groepnamen geven.
  25. Autofit header
    http://jquery-textfill.github.io/
    https://github.com/jquery-textfill/jquery-textfill
Nice to have
  1. Titel en omschrijving toevoegen aan de file upload
Workflow ORM
  1. Recognize bar
  2. Find Tempo
  3. (Find Title)
  4. Find key
  5. (Find accidentals)
  6. Find notes
    • posistion
    • duration
  7. (Extra tekens)
  8. (Text)

1. Upload images

Drag and drop a file here or click

Ooops, something wrong happended.

Successfully uploaded image.

    2. Select training image

    Card image cap

    Card title

    Some quick example text to build on the card title and make up the bulk of the card's content.

    Button

    3. Create training data

    images/download.png

    Select other Reload
    Upload

    4. Show training data

    5. Get images from server

    6. Crowdsourcing

    Classify this image

    7. Show results

                        Show all classes with number of images
                    show number of responses per class

                    choose category
                    random show image

    8. Main take-aways

    More information on Machine Learning
    • https://www.tensorflow.org/js
    • https://ml5js.org/reference/api-charRNN/
    • OMR-Datasets
    Other resources
    • https://www.html5rocks.com/en/tutorials/webgl/typed_arrays/
    • https://towardsdatascience.com/active-learning-on-mnist-saving-on-labeling-f3971994c7ba
    • https://rubikscode.net/2019/03/25/image-classification-with-tensorflow-js/
    • https://medium.com/@ashok.tankala/build-the-mnist-model-with-your-own-handwritten-digits-using-tensorflow-keras-and-python-f8ec9f871fd3
    Computer Vision
    • https://colab.research.google.com/github/lmoroney/mlday-tokyo/blob/master/Lab2-Computer-Vision.ipynb#scrollTo=lR-a4phzaQCF
    • https://github.com/zalandoresearch/fashion-mnist
    • http://yann.lecun.com/exdb/mnist/
    More information on music players Guitar lessons Other