Sie sind auf Seite 1von 8

TPR2251 Pattern Recognition

Assignment
1.

Introduction
The objective of this project is for you to gain experience in designing a pattern recognition
system for solving practical problems by applying the knowledge and tools learned in this course.
You need to develop a system which takes an image as input and recognizes its class. You can
use the Principal Component Analysis (PCA) method learnt in the class for this project. However,
you are encouraged to investigate advanced topics like SVM and Bayesian classifier as well. This
will be treated as an extra effort you put in this project which will give you extra credit. Read the
instructions below carefully. The mark breakdown for each activity is given.

2.

Grouping
You can form your own team. Maximum of 4 members in a team.

3.

Project Details

3.1

Data Collection (4%)


You are going to take your own photos for the project. You are not allowed to use the images
downloaded from the Internet. You must include time and date in your photos as a proof of
your work. Find out how to set up date and time in your digital cameras.
As you are going to develop a fully functional and robust object recognition system, your system
must be able to recognize objects with different variations including changes in lighting and
viewing angles. Your system should also be able to recognize the objects through different times
of the day (e.g. morning, afternoon, sunset, night). Some sample images of the famous Christ
Church with different variations are shown in Figure 1 and 2.

Figure 1. Images of Christ Church taken from different angles.

Figure 1. Images of Christ Church taken at different times of a day

Please follow the following steps to capture your photos:


i.
ii.

iii.

3.2

In a photo-taking session, take at least 10 photos for each object which comprise of view
angle variations (e.g. front view, side view, back view).
Have a photo-taking session every week, for a minimum of 3 weeks. This is to make sure
your object recognition system can work for object changes across time. Please turn the
Date and Time function to On in your camera. Make sure the date and time you take
the photo are shown on the photos that you submit together with your project.
You can put in other variations such as having a person to stand in front of the building
and block some part of it (this is known as occlusion) during your photo-taking sessions.
This depends on your creativity and extra credits will be given to you for the effort you
have put in to capture the photos.

Pre-processing (2%)
Pre-process the images. You can segment your objects manually from the images (e.g. use the
crop function). However, you can also use automated technique to segment your objects. This is
considered your extra effort for this project.

3.3

Feature Extraction (2%)


You can either use the Principal Component Analysis (PCA) or Fisher Discriminant Analysis (FDA)
techniques learnt in the class for this purpose.

3.4

Classification (3%)
You have a choice to make. The mark distribution for each technique is given below:
Method
Mark
Euclidean Distance
1
K-Nearest Neighbor (K-NN)
2
Neural Networks
3
Give the detail on how you classify your objects using the method in your report. Show the code
in your explanation.

3.5

Interface (2%)
You can have a user interface to control your system. The mark is given depending on how
attractive and user-friendly your interface is.

3.6

Report (2%)
Your report should include (but not limited to) the following sections:
a)
b)

c)
d)

e)
f)
g)
3.7

Introduction
Description of design and implementation: Block diagram/flow chart of your system,
methods used for segmentation, feature extraction and classification, highlights of
special features (extra methods) in the project.
Description of your input data (the photos you have taken)
Experiment setup: Data collection (e.g. number of images collected, camera
specifications, image resolution), training and testing images (it is recommended to use
60% of the samples as training images and 40% as testing images).
Performance evaluation: Shown your results in table or graph.
Discussion and conclusions: What you have learnt from this project
References

Presentation (3%)
You are required to present your system to the class. Your presentation should be about 15
minutes. You can refer to the following documents for how to give a good talk:
a)
b)

HowtoPrepareYourPresentation3.pdf
How to Give a Bad Talk.docx

Your presentation will be based on peer-review. In other words, all of your fellow classmates
will assess your presentation (including yourself). Just think of American Idol, you get votes from
everyone. Refer to the documents for the evaluation criteria:
(i)
(ii)

Grading_courseMates.pdf: Evaluation by fellow classmates.


Grading_groupMembers.pdf: Evaluation by team members.

The average score will be taken as your final mark.

3.8

Extra Effort (2%)


This section will be dedicated to whatever extra effort you have put into this project which is not
mentioned in this document. It can either be some special enhancement you put in to improve
your algorithm or extra feature for your object recognition program.

4.

Due Date
The report submission due is on 5 February 2016, 5.00 pm. One mark will be deducted for each
overdue day. Items to be submitted are:
(a)
(b)
(c)

Report
Source code
Dataset (original + processed)

The presentation will be in Week 13.

Appendix A
1. Petrol Station Recognition System
E.g. Shell, Caltex, Petronas

2. Shopping Mall Recognition System


E.g. AEON, Mahkota Parade, Pahlawan Square

3. Bank Building Recognition System


E.g. Maybank, Public Bank, CIMB

4. Car Showroom Recognition System


E.g. Mercedes-Benz, Lexus, Volkswagen Showrooms in Melaka

5. Universities Recognition System


E.g. Multimedia University, Universiti Teknikal Malaysia, Melaka Learning Centre Open
University

6. Kindergarten Recognition System


E.g. 9. Kindergarten around Melaka

7. Hospital Recognition System


E.g. Pantai Hospital Melaka, Mahkota Medical Center, Putra Hospital Melaka

8. Factory Recognition System


E.g. Infineon, Texus Instrument, AUO Sunpower

9. School Recognition System


E.g. Schools around Melaka area

10. Mamak Foods Recognition System


E.g. Roti Canai, Mee Goreng, Nasi Lemak

11. Sport Balls Recognition System


E.g. Shuttlecock, Ping Pong ball, Football

12. Music Instruments Recognition System


E.g. Guitar, Piano, Saxophone

Das könnte Ihnen auch gefallen