AmerGong / NYPL_HTRModel_NYU_ITP2021summer

NYPL-HTRModel-NYU-ITP2021summer

Handwritten Text Recognition (HTR) for NYPL

NYU-Information Technology Projects-2021 Summer-NYPL Group

Group memeber: Anci Hu, Ke Shi, Vipul Goyal, Yuze Gong

Handwritten Text Recognition (HTR) system implemented using TensorFlow 2.x and trained on the NYPL offline HTR datasets. This Neural Network model recognizes the text contained in the images of segmented texts lines.

Data partitioning (train, validation, test) was performed following the methodology of each dataset. The project implemented the HTRModel abstraction model (inspired by CTCModel) as a way to facilitate the development of HTR systems.

This model is an improved version of HTRModel 's NYPL

Datasets supported

a. NYPL

b. Bentham

c. IAM

d. Rimes

e. Saint Gall

f. Washington

Requirements

• Python 3.x
• OpenCV 4.x
• editdistance
• TensorFlow 2.x

Sample

NYPL sample with default parameters in the tutorial file.

1. Preprocessed image (network input)
2. TE_L: Ground Truth Text (label)
3. TE_P: Predicted text (network output)

Tutorial

Step1.Clone the project to local

git clone https://github.com/AmerGong/NYPL-HTRModel-NYU-ITP2021summer.git

Step2.Access to date

A Jupyter Notebook is available to access to NYPL data, check out the NYPL_API.

Step3.Make the NYPL dataset

NYPL_dataset
We made the data set ourselves. Use Readcoop to get the ground truth. And use the picture editing tool to cut the images into lines. Process the cut pictures through the imageresize.py file.

###Step4.Open the project locally Use any code editor you like. It is strongly recommended to use PyCharm.

###Step5.

A Jupyter Notebook is available to demo run, check out the tutorial.