This Notebook aims at introducing ImageDataGenerator, and show how it can be used to train a model to recongise sign language in a set of images. You will need to download the dataset from Kaggle https://www.kaggle.com/datamunge/sign-language-mnist and save it in your local drive. The use of ImageDataGenerator enable us to do data augmentation which is an essential task in many machine learnign algorithms. For more details see keras docs https://keras.io/preprocessing/image/.
You can download the Notebook eithe by clicking Clone and Download button (top right) link or using command line:
$ git clone https://github.com/heyad/DataAug.git
- Add more conv/pooling layers
- Change the 'ImageDataGenerator' and see the impact on the performance of the model
train_datagen = ImageDataGenerator(
rescale = 1./255.,
rotation_range=5,
width_shift_range=0.13,
height_shift_range=0.3,
fill_mode='nearest')
