What needs to change about training with ifN-ENIT datasets？

[CHENWEI-1213]

Hi, I now want to follow the training in your paper, what changes should I make if I want to train on the above data set

[msfasha]

The model presented in the paper uses a single file for storing image files i.e. the .bin file.

This was needed because the number of training samples was large i.e. few millions. Also, loading a single large into the cloud (Google Colab) file was more efficient because the presented model was tested using numerous datasets that were generated using the Java tool provided in the project.

Therefore the binary format assisted in better usage and facilitated the loading of sample batches into memory.

For single word samples. you will need to modify the code in the model that loads the binary file, rather, you will need to load
the sample images from a directory/folder into a list and use that list to train the model. This should not be difficult, the model will be mostly the same, but you need to be careful about image sizes when you feed to the model.
In addition, you need to make sure that the labels data is compatible with what the model is expecting.

Get back to me if you are stuck or need any assistance.