Characters Recognition using AlexNet Convolutional Neural Network

Prerequisites

• Python 3
• Keras 2.2.4
• Tensorflow 1.13.1
• OpenCV 4
• Numpy

Usage

Uncompress the file English.tar.gz into a folder and update the constant DATASET_PATH in train.py accordingly.

To train the model:

python train.py

The model will be saved in the path specified by the constant MODEL_PATH. The model's file name is "trained_model".

Model performance:

train accuracy: 97.29%
validation accuracy: 95.46%

To test the trained model on an image, just run:

python predict.py --image name_of_the_image_here

Full tutorial is available at http://dangminhthang.com/computer-vision/character-recognition-using-alexnet/