Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit

Project Title

NOTE: Delete the yellow highlighted regions and ADD YOUR TEXT

1 Introduction

DEADLINE: Sept 2^nd, 2023  – 5pm

- General Problem Description

- What is unique on this project? Database ▪ Algorithms used ▪ Solution/Approach ▪ Did you implement a Good/Best Paper (CVPR)? etc.

- What did “you” do? – Contributions:

o DBs (did you collect/generate it? Do you use an existing DB? Do you extend an existing DB with your own data e.g., for testing)?

o Same for code

o Is it your own?

o Are you using toolboxes provided online (e.g., specific MATLAB DB) – where/cite it?

o OpenCV?

o Are you using someone else’s code from a paper online as is or also improved it or combined different algorithms (your approach)? etc.

2 Literature Review

DEADLINE: SEPT 20^th, 2023  – 5pm

- Find no less than 10 papers most relevant to what you are trying to work on:

Example format expected for the review of your papers:

- Kong and Zhang [1] proposed several approaches to reduce eyelash and reflection noise on iris segmentation. The eyelash detection model was based on separable eyelash condition, non-informative condition and connective criterion. The reflection detection was based on strong and weak reflection tests. They concluded that the proposed noise reduction methods are effective in terms of detecting eyelash and reflections.

- Huang, Wang, Tan and Cui [2] explored an iris segmentation method using phase congruency. Iris is firstly localized and normalized to a rectangular block with fixed size. Then edge information is extracted based on phase congruency by a bank of Log-Gabor filters. Finally, the edge information of noise region is infused to iris segmentation, including eyelash, eyelid, reflection and pupil.

- Liu, Bowyer and Flynn [3] developed an improved segmentation algorithm based on Masek’s algorithm. The detection order is reversed by segmenting the pupil boundary first. Then some edge points not from the iris boundary are eliminated in order to find correct iris boundary using Hough transform. Instead of each edge point voting in all directions, the proposed algorithm only let each edge point vote for 30 degrees on each side of local normal direction. Finally, hypothesize is performed to ensure the accuracy of the algorithm.

3 Methodology

DEADLINE: OCT 10^th, 2023  – 5pm

Tools:

- Programming language (e.g., MATLAB, Open CV) and toolboxes use

- Brief Description of the Algorithms used (e.g., for feature extraction, classification or clustering, statistics etc.)

- Description of the “Overall Approach”: flow chart of the process (for example: input image, pre-processing, feature extraction, classification…)

- Description of why this approach is useful and important in this application you work (e.g., wood classification)

4 Experiments and Results

DEADLINE:

OCT 31^st, 2023  – you can then,

- continue to update Section 4 in the follow up report by NOV 30^th, 2023

- Description of the DB used, #images, samples etc.

- Description of the experiments performed

- Error Analysis (e.g., effect of image size in classification performance)

5 FINAL REPORT - Conclusions

DEADLINE: NOV 30^th, 2023

- Additional Experiments and Refinement of your algorithm’s efficiency.

- Additional Testing and Evaluation

- What did your experiments reveal for the efficiency of your algorithm etc.

6 Appendix

DEADLINE: NOV 30^th, 2023

- Code………...(needs to be submitted last week before finals) - CD

- Demo………..(needs to be presented last week before finals)