search

clovaai / deep-text-recognition-benchmark

Text recognition (optical character recognition) with deep learning methods, ICCV 2019
Apache License 2.0
3.77k stars 1.11k forks source link

Normalization about own data #308

Open BENHSU0723 opened 3 years ago

BENHSU0723 commented 3 years ago

hello, i'm try to train on my Chinese datasets. During the data processing , i have some confusion below.

  1. Should i Normalize the data into same size to get better result??
  2. The textflow in datasets combined with vertical and horizontal , should i Rotate them into same direction ?

by the way, i would choose TRBA (TPS-ResNet-BiLSTM-Attn) for train.

i don't have too much base knowledge about deep learning, thank for reply !!