ICR (Intelligent Character Recognition)

NOTE: This is a very granular level implementation of the ICR for Uppercase Alphabets, thus it can be used to be implemented in projects with ease.

Input:

Output:

Library	Installaion
opencv	pip install opencv-python
tensorflow	pip install tensorflow
numpy	pip install numpy

To run the model on your image data, just provide the path of the file in the ICR.py file and change the following line

img = cv2.imread('path/to/your/file.extension')

and then run the following code in the cmd or terminal python ICR.py you can see your result in the main directory of the project with name contoured1.jpg it can also be renamed in the same file by going to the line

cv2.imwrite('filename.extension', img)

---

If you wish to use the pretrained weights they are saved by name weights.h5, just load them using the load_model function from tf.keras.models

# Example
model = load_model('weights.h5')

---

If you wish to train the same neural network on your own dataset, make the following changes in the following lines in either neural_model.py.

On line 10

CATEGORIES = ['Categories', 'That', 'Task', 'Needs']

On line 11

DATADIR = 'Path/to/your/dataset'

On line 55

model.add(Dense('No. of categories you added in categories list'))

On line 68

image = cv2.imread('path/to/your/test/data', 0)

And then simply run the run the model by python neural_model.py

Note:

The file structure for training data should be
Dataset
   |- Category 1
          |- Data for category 1
   |- Category 2
          |- Data for category 1
   :
   :
   :
   |- Category n
          |- Data for category n

If not please make the necessary changes in the script to load the data as per your need