Paid Projects , Project ,

Data Exploration and Building basic model for Text Recognition

In this project, we’ll explore and understand the IAM dataset, which contains handwritten text images and their parallel digital text and English-Hindi parallel corpus for machine translation models. We’ll start with the base model using CRNN architecture for the offline OCR. First, we’ll build an OCR model for detecting and extracting words in the images and then move on to extracting sentences or lines in the Images. 

What will you Learn in the Project?

  1. How to load IAM text and images data for model building 
  2. Understanding the IAM data structure 
  3. How to build deep learning models with CRNN architecture
  4. Understanding CTC for loss for text corrections 
  5. Training and evaluating OCR model on the IAM dataset

Tools & Technologies Used

  1. Google Colab 
  2. Pandas 
  3. Numpy 
  4. Matplotlib
  5. Tensorflow [Keras]

Tasks Performed

Task-1: Explore and Understand the IAM(words) dataset

Task-2: Explore and Understand the IAM(lines) dataset

Task-3: Explore and Understand the English-Hindi parallel corpus for translation

Task-4: Create a base offline OCR model using CRNN for IAM(words)

Task-5: Create a base offline OCR model using CRNN for IAM(lines)

Not Enrolled
or 999₹ 9999
91% off

Skills you will develop

Read and write data from text files and export it into a pandas data frame using Python

Load the dataset into memory using TensorFlow data API

Create distortion-free images with padding

Create a CTC loss class as a model layer

Create Edit Distance Callback class for calculating edit distance between predicted text and the actual label

Make predictions using a trained Deep Learning model and visualize those results and actual images

Video format is not supported, use Youtube video or MP4 format.
Share with Friends and earn points!!