CSC508 – Computer Vision – Fall 2003



CSC508 – Computer Vision – Fall 2010

Final Exam

Due: Thursday November 11, 2010

You are given a color image of 1, 2, or 3 dice. Some features of the images include:

1. Perspective is present, but minimal. That is, the images show primarily the upward face and small amounts of the side faces.

2. Background color is of high contrast to the dice, but unknown a priori.

3. Lighting varies and is unknown a priori.

4. Spatial placement of the dice relative to one another (when multiple dice are present) is unknown a priori.

5. Number of dice in any given scene is unknown a priori.

6. Orientation (rotation) of the dice is unknown a priori.

7. Images do not contain any other items.

8. All images are 24 bit RGB color, 640x640 pixels.

Figures 1 – 5 show some example images each containing 3 dice.

[pic]

Figure 1 – Three dice, pip count of 14, arbitrary layout.

[pic]

Figure 2 – Three dice, pip count of 14, arbitrary layout.

[pic]

Figure 3 – Three dice, pip count of 14, touching.

[pic]

Figure 4 – Three dice, pip count of 14, occlusion.

[pic]

Figure 5 – Three dice, pip count of 14, severe lighting.

Your job is to design an automated computer vision system consisting of low, mid, and high-level modules that reports

1. The number of dice in the scene

2. The total pip count of the upward faces

The input to the system is an arbitrary image. Any model information you require must be built into the system. That is, you cannot use a model designed specifically for a given input image. The output from the system is the number of dice and the pip count as described above.

Specify the algorithms (pre-processing, low level, mid level, and high level) that will make up your system. For each algorithm state:

1. Purpose for inclusion of the algorithm in the system

2. Inputs to the algorithm

3. Outputs from the algorithm

4. Basic control flow of the algorithm

5. Basic data structures used by the algorithm

6. Demonstration of the algorithm if you have software available (from an earlier project or a tool such as ImageJ.)

7. Alternatives (a back-up plan) for this algorithm

Wherever possible, demonstrate your algorithm selections in software, using the provided input images (see note below.) Some images will be conducive to your algorithmic choices, others will not. In those cases where a particular image does not work well with your selected algorithm(s) you must describe the problem(s) encountered and postulate viable solutions. Where you don’t actually implement code, make sure your descriptions are clear (consider flow charts or pseudo code along with your textual description.)

Specify the interconnections between algorithms including what data will get passed from one algorithm to the next.

Note: I expect this to be a very difficult problem to solve and the required algorithms and software extremely difficult to get working, especially in the toughest image cases. I’m not looking for a fully implemented, tightly coupled, working software system. I am looking for your understanding of the algorithms we’ve studied, utilization of the software you wrote to explore those algorithms, how you might use algorithms/software that we did not write ourselves (e.g. you may use operations in ImageJ or any other tool available), and how you might link the algorithms together into a working system.

Grading will be based on:

• Completeness of design – 25%

• Feasibility of design – 25%

• Appropriate usage of algorithms – 40%

• Presentation (clarity, layout, grammar, etc.) – 10%

To date, all assignments have been counted equally. This will be weighted equivalent to four assignments.

You will turn in a single document (report) including processed images (contained within the document…do not hand in multiple files!) and a description as to how you processed them – e.g. did you use your code or code provided through a tool such as ImageJ. DO NOT turn in source code unless you include it in the paper. The due date is the date of our final exam. Late papers will not be accepted. You may use the class time during the final exam to ask questions and finish things off if necessary. Note that questions must be of a generic nature. I will not tell you how to solve the problem. I will be here for the entire final class period but will not be delivering any lecture material. THIS IS NOT A GROUP PROJECT! You are expected to turn in your own work.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download