I want to make my own OCR with machine learning

I want to make my own OCR from scratch. I am not even well acquainted to machine learning practices in detail. Can someone suggest me where to start from. My goal is that I give an image to obtain the text from it. Any help is appreciated! Thanks.