mwanjajoel / clothia Goto Github PK

When splitting the dataset it is always advisable to split it into train, validation, and test set. That is if you are not using cross-validation. I do not want to go into the details of cross-validation since it's your first project but you will meet it along the way.
Your dataset is very huge (60k images). You can create a small validation set (it could be about >=5k images). This one helps you monitor the performance of the model during training i.e if the model is overfitting or underfitting. This information will let you know when to halt the training process.

More information about the validation set can be found here https://machinelearningmastery.com/difference-test-validation-datasets/
https://www.kdnuggets.com/2017/11/create-good-validation-set.html

You trained your model for only 5 epochs and it exhibited a good performance (accuracy kept on increasing and the loss kept on decreasing). This shows a great training process. It would be interesting to see what happens when you gradually increase the number of training epochs because your results indicate that you have not yet reached the model's optimal performance.