This work dedicates research on applying Deep Learning to the practice of EAN-13 barcode recognition. In the work, we firstly synthesize the data set for training based on publicly accessible Flickr data set. We then train and adapt object detector based on the synthesized data set. To read the digits in the EAN-13 barcode, we propose to use both classification and regression model to decode the information. Experimental results show that this is an applicable way for barcode recognition and we wish it to inspire future works.