Author avatar

Liam

02 Apr, 2016

blogpost banner

TensorFlow Udacity 1_notmnist - Part 3

Summary of 1_notmnist

Basically 1_notmnist is to learn how to display data in Jupyter Notebook. Besides, it also let us know on sklearn - a python machine library - so that we can then compare with TensorFlow. This is the exact ipynb file at Tensorflow Github Repo.

Notice

This is as a form of sharing and discuss on better way to solve 1_notmnist problem. Do not copy and paste directly as it does not help on improving yourself + the answer is not optimized.

The entire series of TensorFlow Udacity can be found at tensorflow-udacity tag

Solving Problem 3

len_dict = {}
for folder in folder_names:
    t = pickle.load(open(dir_name + "/" + folder + ".pickle", "r"))
    len_dict[folder] = len(t)
print(len_dict)

png

Comment

Since the image for each class is roughly around 52911, therefore we can say that the data is balanced. Though more appropiate way is to calculate the standard deviation :)