GitHub - shrey-bansal/ALPNR: Automatic License Plate Number Recognition

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 3 Commits
.idea		.idea
__pycache__		__pycache__
mnist		mnist
Digit_Segregation.py		Digit_Segregation.py
Image_Preprocess.py		Image_Preprocess.py
License_Plate_Detection.py		License_Plate_Detection.py
License_Plate_Detection.pyc		License_Plate_Detection.pyc
Main.py		Main.py
Plate_Detection.py		Plate_Detection.py
README		README
a.out		a.out
car.jpg		car.jpg
cars.jpg		cars.jpg
digit_clustering.py		digit_clustering.py
model-neural-network.dat		model-neural-network.dat
testing-report.dat		testing-report.dat
testing_nn		testing_nn
testing_nn.cpp		testing_nn.cpp
testing_nn.py		testing_nn.py
testing_nn.zip		testing_nn.zip
training-report.dat		training-report.dat
training_nn		training_nn
training_nn.cpp		training_nn.cpp
training_nn.py		training_nn.py

Repository files navigation

Software: Artificial Neural Network for MNIST database (C++)
Author: Hy Truong Son
Major: BSc. Computer Science
Class: 2013 - 2016
Institution: Eotvos Lorand University
Email: sonpascal93@gmail.com
Website: http://people.inf.elte.hu/hytruongson/
Copyright 2015 (c). All rights reserved.

Overall
-------
Neural Network implementation in C++ running for MNIST database.

Structure
---------
File training_nn.cpp: the code for training a neural network
File testing_nn.cpp: the code for testing a trained neural network
File model-neural-network.dat: contains the weights of the neural network
File training-report.dat, testing-report.dat: report files, saving results of training and testing
Folder ~/mnist/: MNIST database

Note: model-neural-network.dat is the input for teting process (testing_nn.cpp)

Usage
-----
* Compile:
$ g++ training_nn.cpp -o training_nn
$ g++ testing_nn.cpp -o testing_nn

* Training:
$ ./training_nn
**************************************************
*** Training Neural Network for MNIST database ***
**************************************************

No. input neurons: 784
No. hidden neurons: 128
No. output neurons: 10

No. iterations: 512
Learning rate: 0.001
Momentum: 0.9
Epsilon: 0.001

Training image data: mnist/train-images.idx3-ubyte
Training label data: mnist/train-labels.idx1-ubyte
No. training sample: 60000

Sample 1
Image:
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000001111111111110000
0000000011111111111111110000
0000000111111111111111100000
0000000111111111110000000000
0000000011111110110000000000
0000000001111100000000000000
0000000000011110000000000000
0000000000011110000000000000
0000000000001111110000000000
0000000000000111111000000000
0000000000000011111100000000
0000000000000001111100000000
0000000000000000011110000000
0000000000000011111110000000
0000000000001111111100000000
0000000000111111111000000000
0000000011111111110000000000
0000001111111111000000000000
0000111111111100000000000000
0000111111110000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
Label: 5
No. iterations: 512
Error: 0.009284

Sample 2
Image:
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000001111100000000
0000000000000011111100000000
0000000000000111111111000000
0000000000011111111111000000
0000000000011111111111000000
0000000000111111111111000000
0000000001111111110011100000
0000000011111100000011100000
0000000111111100000011100000
0000000111100000000011100000
0000000111000000000011100000
0000001111000000000011100000
0000001111000000001111100000
0000001110000000011111000000
0000001110000000111100000000
0000001110000001111000000000
0000001111111111111000000000
0000001111111111100000000000
0000001111111110000000000000
0000000111111100000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
Label: 0
No. iterations: 512
Error: 0.007427

...

* Testing:
$ ./testing_nn 
*************************************************
*** Testing Neural Network for MNIST database ***
*************************************************

No. input neurons: 784
No. hidden neurons: 128
No. output neurons: 10

Testing image data: mnist/t10k-images.idx3-ubyte
Testing label data: mnist/t10k-labels.idx1-ubyte
No. testing sample: 10000

Sample 1
Error: 0.000000
Classification: YES. Label = 7. Predict = 7

...

Sample 9999
Error: 0.000001
Classification: YES. Label = 5. Predict = 5

Sample 10000
Error: 0.000002
Classification: YES. Label = 6. Predict = 6

Number of correct samples: 9440 / 10000
Accuracy: 94.40