Bangla numeral recognition using multilayer feed forward, Computer Engineering

Assignment Help:

Assignment 4: Handwritten Bangla Numeral Recognition using Multilayer Feed Forward Neural Network.

In this assignment, you will design a multi layer feed forward neural network (MLP) classifier that will be used to recognize handwritten Bangla numerals. There are ten numerals in Bangla alphabet. The data set that will be used consists of image samples of handwritten Bangla numerals which were found by asking people of various professions to write down the numerals in paper and then scanning the paper to images. The data set consists of a training and test set where in each set there are 25 samples for each numeral. Therefore, in each dataset, there are 10X25 = 250 samples. You will have to use MATLAB image processing functions to do this assignment.

1.       Download the training and test data from my website.

2.       This machine learning task involves four key phases.

a.       Phase 1- Preprocessing:  

 i.      In this step, the images samples will be read using MATLAB function imread. The raw data that will be returned by imread is the pixel-by-pixel luminosity. So, if image size is 50x50, the imread function a 50x50 matrix, where each cell denotes the intensity of the image in the corresponding pixel. Intensity is between 0 and 1. This is what called grayscale image.

 ii.      Since, each image is not of the same size; there we have to resize them to a uniform size. In this step, use the MATLAB function imresize to resize each image matrix into a uniform size of 40x40.

 iii.      After resizing, you have to convert the grayscale image to a binary image (where each pixel will be either 0 or 1). This will be done by using graythresh and im2bw function.

 iv.      [You can thin the image using bwmorph function. Although this step is not necessary for this assignment, however you can experiment with it, whether thinning gives improvement in performance

b.       Phase 2- Feature extraction:

i.      In this phase, you will extract features. In this assignment, feature will correspond to the average intensity of a 5x5 size box centered at each pixel. All such average intensities will form the feature vector.

c.        Phase 3- Training the MLP:

 i.      Train the back propagation neural network using training data set.

d.       Phase 4- Testing: Test the learned neural network and find accuracy measures.

3.       Perform experiments by varying:

a.       Number of hidden layers: Experiment with 5, 10, 25, 50.

b.       Number of iterations: Experiment with 100, 300, 500, 1000.

c.        Size of the box: Experiment with 3x3, 5x5, 9x9, 15x15.

d.       In each of the above experiments, take average of 25 runs.

4.       If you want to determine the effect of factors stated in (3), then run 4x4x4 = 64 experiments. Comment on the results.

5.       Write a report commenting on the experiment results. State your findings explicitly. These findings will give you better marks.


Related Discussions:- Bangla numeral recognition using multilayer feed forward

Structural hazards - computer architecture, Structural hazards - computer a...

Structural hazards - computer architecture: A structural hazard takes place when a part of the processor's hardware is required by 2 or more than two instructions at the same

Ida* search - artificial intelligence, IDA* Search - artificial intelligenc...

IDA* Search - artificial intelligence: A* search is a sophisticated and successful search strategy. In fact, a problem with A* search is that it must keep all states in its me

What is thread, What is Thread? A thread, sometimes termed a lightweigh...

What is Thread? A thread, sometimes termed a lightweight process (LWP), is a fundamental unit of CPU utilization; this comprises a thread ID, a register set, a program counter

Illustrate what is a astrophysics, Q. Illustrate what is a astrophysics? ...

Q. Illustrate what is a astrophysics? Answer:- Examine of galaxies, stars and the creation of the universe including predictions about how it will proceed from here.

What is mmx technology, What is MMX Technology MMX Technology: MMX (M...

What is MMX Technology MMX Technology: MMX (Multimedia extensions) technology adds 57 new instructions to instruction set of Pentium - 4 microprocessors. MMX technology also

Find how many bits of adc are required to get an resolution, Find how many ...

Find how many bits of ADC are required to get an resolution of 0.5 mV if the maximum full scale voltage is 10 V Ans. Given Resolution=.5mV Full scale output=+10v %resolut

What is database integration, What is database integration? Database in...

What is database integration? Database integration is the ability to give user-friendly and cost-effective software solutions for data infrastructure management by the interfac

Texture - elements of composition, Texture Texture is how artists show...

Texture Texture is how artists show what an object would feel like to touch. This can have a dramatic effect on how an image is perceived. Imagine for instance an image that h

Explain difference between scanning and parsing, Explain difference between...

Explain difference between Scanning and parsing. Scanning and parsing: Scanning process is the process of recognizing the lexical components in a source string whereas parsin

Feature Extraction and clustering, Hi, I have a CSV file that has numbers ...

Hi, I have a CSV file that has numbers data set. The data set contains images of handwritten digits. Recognizing handwritten digits is already a mature technology By using R code

Write Your Message!

Captcha
Free Assignment Quote

Assured A++ Grade

Get guaranteed satisfaction & time on delivery in every assignment order you paid with us! We ensure premium quality solution document along with free turntin report!

All rights reserved! Copyrights ©2019-2020 ExpertsMind IT Educational Pvt Ltd