The Clasificacion Basica: Predecir una imagen de moda | TensorFlow Core tutorial has been tremendously helpful for learning the steps for classification. The mnist clothing example (Classify images of clothing) uses
fashion_mnist = tf.keras.datasets.fashion_mnist
as the test dataset. This is further flattened from 28*28 to 784. I see that related blogs exist, but I could not find an answer to my question, which is how to substitute my data for the mnist data. I have generated and exported the mnist data into a csv file (needed to use pd.DataFrame and df_to.csv), but reversing the order for input (pd.read_csv) , most of the features of the above fashion_mnist dataset are lost. Can anyone provide a recipe for exporting the mnist dataset as any ascii format, such that it can be re-read as input for classification? To re-state, my goal is to use the keras classification steps on my own data.
Thanks in advance
Hi @david_covell, Could you please let me know if my understanding is correct or not, you want to use the mnist flatten csv data for training the model.
reversing in the sense after flattening the data from 28X28 to 784, are you considering the input in the reverse order i.e. 783-0.
you want to convert the whole dataset to ASCII values or only futures or only labels. Thank You.
My goal is to export the mnist dataset (which is loaded with either MNIST or load_mnist). Then edit the content of the exported csv file, and reload the edited data. So far, I have not been able to mimic the format achieved when using the existing utilities to load the mnist data. Although this seems like a trivial task, it is necessary to carry forward the various codes available for mnist analysis when applied to alternative data. In other words, most of the tutorials related to mnist data appear to be teathered to the format used for loading the mnist data. Accordingly, loading your own data does not seem to be straight-forward. I have tried loading the titanic data and the abalone data (see โหลดข้อมูล CSV | TensorFlow Core). without success.
Any advice would be helpful.