search

githubharald / CTCWordBeamSearch

Connectionist Temporal Classification (CTC) decoder with dictionary and language model.
https://towardsdatascience.com/b051d28f3d2e
MIT License
557 stars 160 forks source link

Result of Paper published #55

Closed jjsr closed 3 years ago

jjsr commented 3 years ago

Hello Sir, While going through CTC WBS paper, kindly please tell while writing these results , table 1) Have you used actual line of images available with Dataset official zip files(IAM and Bentham) , or Take a Page level Images then applied word and line segmentation (which ever applicable) and then use these word or line images to calculate CER/WER for the results to be put on result table

2) What Train, Test and Validation split have you used for IAM (As per simpleHTR Is it 0.95 and 0.05 for train and validation what about test??) for Bentham dataset Test Test and Validation is given so you may have used that

Thanking in Advance !!!

githubharald commented 3 years ago
  1. yes, took the line images from both datasets
  2. see https://fki.tic.heia-fr.ch/databases/iam-handwriting-database under Tasks, there you can find the split between into training and testing set
jjsr commented 3 years ago

Thanks Sir ,