How to train using the original lmdb dataset from this repo but also using a custom dataset?

[CharlyJazz]

I need to merge the lmdb dataset that this repo use in the instructions with my dataset? Because when I use only my dataset the OCR is not as good as the OCR trained with the dataset created for this repo. Thanks!

[AdamKheire]

@CharlyJazz did you get the solution

[CharlyJazz]

There is a class ConcatenatedDataset, you can put your dataset side by side of the others datasets but keep in mind that the the are sorted by folder name, and it gonna take a long time training on the originals lmdb, so your dataset should be first, to make sure it gonna be use in the training session first. I start my dataset name with "a"