CRNN

paper:An End-to-End Trainable Neural Network for Image-based Sequence Recognition and Its Application to Scene Text Recognition


1. create your own dataset

Process different text texts, and store image path and label label in LMDB database in one-to-one correspondence. Here, the text format I processed is as follows:

./9/7/186_densely_20626.jpg 20626
./9/7/185_acidic_759.jpg 759
...

Extract useful information from text 186_densely_20626.jpg , build target path and correspond to label: densely


2. dataset

Remove the following characters when reading LMDB data

b'******' ——→ '******'

3.train

GitHub

View Github