March 1st, Monday
Today I ran the neural net for hiragana image recognition model in google colab. It worked a lot better than any of my previous attempts reaching a accuracy of 98ish percent. but it was kind of slow and took about 20 minutes to train.
March 3rd, Wednesday
today I tried to convert the hiragana model from the .h5 file format to the .mlmodel format. However I kept on getting errors that said that "the layer was not interable". I'm not really sure what is going on.
Summary
this week i made major progress finally getting the model to train. I think more than how to do machine learning I am learning how to troubleshoot. As soon as I get this model to work I can repeat the process for the katakana and kanji models.
Comments