Assignment due Monday, May 11, 2026, by 11:59pm

Computer Vision System Design, Implementation, and Evaluation

Submission:

Primary submission: Final Report (PDF, max 10 pages excluding references, including video link (with proper sharing settings so that the video is accessible))

Secondary submission: Zip file containing Jupyter Notebook(s) (.ipynb) and/or Python scripts (Important: you should avoid including large hidden files while copying your code (avoid direct copy of PyCharm projects, which may include large hidden folders)). Do not include the video file in this package.

1. Overview

The Final Project is the implementation and evaluation phase of the system proposed in Assignment 2.

Topic Consistency Requirement:

The final project is expected to follow the topic proposed in Assignment 2. If a group decides to change its topic, the report must include a justification statement explaining the reason for the change. In such cases, the group is expected to treat the new topic as a fresh project and perform the necessary design, implementation, and evaluation work so that the project goes well beyond a proposal-level description.

Each group must:

1. Implement both proposed approaches
2. Perform structured evaluation
3. Compare performance across multiple dimensions
4. Analyze runtime complexity and scalability
5. Reflect critically on strengths, weaknesses, and real-world applicability

This project evaluates your ability to apply the full ICS353 pipeline:

• Image formation & preprocessing
• Feature detection and matching
• Motion estimation
• Reconstruction and depth
• Segmentation and recognition
• Evaluation and comparison

ML-based approaches (e.g., Random Forest, AdaBoost) are optional, not required.

2. Learning Outcomes Alignment

This project assesses all course learning outcomes:

#LO1_image: Describe the foundation of image formation and image processing.

#LO2_visionmethods: Compare various computer vision methods used for feature detection, edge computation, motion estimation, reconstruction, and recognition.

#LO3_visiontechniques: Evaluate appropriate computer vision techniques.

#LO4_imageprocessing: Apply image processing techniques to solve computing problems related to object detection and image segmentation for real-world applications.

3. Project Requirements

A. Implementation Requirements

You must implement:

• Sub-team A pipeline
• Sub-team B pipeline
• Complete preprocessing
• Parameter tuning experiments
• Structured evaluation
• Runtime complexity analysis

B. Required Evaluation Dimensions

Your comparison must include:

1. Quantitative accuracy metrics (pick those suit your task)
   • Precision / Recall
   • IoU
   • Matching accuracy
   • Tracking error
   • Depth error
   • F1-score (if applicable)
2. Runtime performance
   • Execution time per image/frame
   • FPS (for video tasks)
3. Robustness analysis (how does your system behave under variation of various factors)
   • Noise sensitivity
   • Illumination variation
   • Parameter sensitivity
4. Failure case analysis
   • Where and why the method fails

4. Final Report Structure (Max 10 Pages)

Your report must include:

1. Title Page

• Project Title
• Group members
• Sub-team assignments
• Course code
• Notebook links
• Submission date

2. Abstract ( 300 words)

• Concise summary including:
• Problem
• Approaches
• Main results
• Key conclusion

3. Introduction ( 5 references)

• Problem context
• Relevance
• Application domain
• Challenges

4. Literature Review ( 5 additional references)

• Summary of related work
• Comparison table of existing solutions
• Strengths and weaknesses of previous approaches

5. Methodology

Dataset Description

• Include a summary table.

Sub-team A Pipeline

• A flowchart/diagram is required
• Preprocessing
• Algorithms
• Justification

Sub-team B Pipeline

• A flowchart/diagram is required
• Preprocessing
• Algorithms
• Justification

Improvement Declaration

Each group must include a short statement explaining how the final system represents a substantial improvement over the proposal submitted in Assignment 2. The statement should clearly describe what was expanded, redesigned, or implemented beyond the proposal stage, including additional experiments, implementation details, evaluation depth, or methodological refinements.

6. Results

Include:

• Quantitative comparison table
• Runtime comparison table
• Robustness tests
• Example outputs (visuals)
• Parameter tuning experiments

Clarity and conciseness are essential.

7. Discussion

You must analyze:

• Why one method performed better
• Trade-offs (accuracy vs runtime)
• Sensitivity analysis
• Real-world feasibility
• Limitations
• Potential improvements

Include:

Curriculum Mapping Table: Technique Used and Session that covered the technique. For example:

• Optical Flow; Session 1820
• Stereo matching; Session 2228
• Feature Detection; Session 78

etc.

8. Conclusion

Summary of findings

Member contributions

3 key learnings for the group

Future work

9. References

Minimum 10 academic references.

Proper citation required.

5. Code and Video Submission Requirements

You must submit clean Jupyter notebook(s) that include the following:

• Organized sections
• Clear comments
• Parameter tuning
• Evaluation metrics
• Output visualization
• Runtime measurement

The notebook must be readable and reproducible.

Project Demonstration Video:

Each group must create a brief demonstration video (maximum 6 minutes) presenting the implemented system, key results, and a short walkthrough of the pipeline. The video should clearly demonstrate the system running and highlight the comparison between the two approaches. The video link must be included in the final report (with proper sharing settings so that it is accessible by the instructor).

7. Academic Integrity

No generative AI-written text

Code must reflect your understanding

You may be asked to explain any part of your code

Attached is the original proposal along with the professors’ comments that need to be made and adjusted to the final.