Ecg signal from ad8232 for Arrhythmia classification project

Question

I intend to make a realtime Arrhythmia classification system, I use ad8232 as a device for getting ecg signal. Here is the project I using https://medium.datadriveninvestor.com/ecg-arrhythmia-classification-using-a-2-d-convolutional-neural-network-33aa586bad67

I use the model he provided, and use his test data(sample.csv), it works properly, but something went wrong when using my own data, but i am not going to ask where I did I went wrong, I would like to know whether my direction is right or wrong.

My question is

1. Is all leads of ecg can be fed into that model to do prediction?
2. He turned ecg signal into 2d image and augmented it, this seems like he is talking about the former step of trainning phase, but does my signal which fed into the model to do prediction will be also transformed in 2d and augmented?
3. Does my signal has to be denoised, because he said signal don't have to be denoised if using 2d convolutional model to predict it
4. The data generate from ad8232 is from 0 to 1024, but its sample data is from -1 < x < 1, If my data is usually between 300 < x < 700, can I process my data like this (x - 500)/1000, make it between -1 < x < 1. Is my method reasonable?

Please tell me which step of my thinking is wrong thank you!