Skip to content

Instantly share code, notes, and snippets.

@tokestermw

tokestermw/data-augmentation-loop.py

Last active December 1, 2019 07:04
Show Gist options
  • Save tokestermw/736dc2b629f9e3a6a53c19d345eccb98 to your computer and use it in GitHub Desktop.
Save tokestermw/736dc2b629f9e3a6a53c19d345eccb98 to your computer and use it in GitHub Desktop.
Download ZIP

Revisions

  1. tokestermw revised this gist Dec 1, 2019. 1 changed file with 17 additions and 13 deletions.
    30 changes: 17 additions & 13 deletions data-augmentation-loop.py
    View file
    Original file line number Diff line number Diff line change
    @@ -2,26 +2,30 @@


    def augmentation_fun(x, augment_by=3):
    return [x + random.random() for i in range(augment_by)]
    # augment the original data point by 3
    return [x + random.random() * 2 - 1 for i in range(augment_by)]


    def train_loop(dataset):
    def train_loop(dataset, do_augment=False):
    # emit one data point at a time
    for x in dataset:
    # ... preprocess, etc.
    yield from augmentation_fun(x)
    if do_augment:
    yield from augmentation_fun(x)
    else:
    yield x


    if __name__ == "__main__":
    dataset = [1, 2, 3]
    for x in train_loop(dataset):
    for x in train_loop(dataset, do_augment=True):
    print(x)
    # 1.8743045582080011
    # 1.1156374021713173
    # 1.9762467469532101
    # 2.658889884418666
    # 2.350529257341205
    # 2.428795321570352
    # 3.9366101605189914
    # 3.6662388553317062
    # 3.285082043706509
    # 1.6910669464085353
    # 0.6755471610373542
    # 1.3990163430931686
    # 1.0574266964192516
    # 1.7474122098401583
    # 2.8820668623545096
    # 3.0684208172249834
    # 2.0069288129912817
    # 3.7746658889260774
  2. tokestermw created this gist Dec 1, 2019.
    27 changes: 27 additions & 0 deletions data-augmentation-loop.py
    View file
    Original file line number Diff line number Diff line change
    @@ -0,0 +1,27 @@
    import random


    def augmentation_fun(x, augment_by=3):
    return [x + random.random() for i in range(augment_by)]


    def train_loop(dataset):
    # emit one data point at a time
    for x in dataset:
    # ... preprocess, etc.
    yield from augmentation_fun(x)


    if __name__ == "__main__":
    dataset = [1, 2, 3]
    for x in train_loop(dataset):
    print(x)
    # 1.8743045582080011
    # 1.1156374021713173
    # 1.9762467469532101
    # 2.658889884418666
    # 2.350529257341205
    # 2.428795321570352
    # 3.9366101605189914
    # 3.6662388553317062
    # 3.285082043706509