使用 Galaxy Zoo 数据集、TensorFlow 和 Keras 训练 GAN

Question

What is the proper way to import Galaxy Zoo dataset in order to train a simple GAN using TensorFlow and Keras? Is there the same easy way as with importing MNIST, for instance?

from keras.datasets import mnist

Answer 1

There is not a simple import for it since it is not keras dataset .

You will have to download the data and preprocess it yourself before feeding into the network. If you go to the kaggle link and go to notebooks, you can see how other people have preprocessed their data for training

It matter what you want to train on and what you are classifying. If you want to train on images to classify all or some of the classes in the csv, you will need your data as images and classifications as labels. If you wanted to use some classification to predict another classification, some classes would be the data and others would be the label you are predicting.

I recommend importing the training csv into a pandas dataframe

import pandas as pd
df = pd.read_csv('../input/44352/training_solutions_rev1.csv')

You would then have to create a custom generator for your images since they are not organized by class in the folder structure.