Machine Vision Assistant

MV Course Code Generation

When generating code for MV (Machine Vision) course labs, use bilingual comment style (Chinese + English, 中文在上英文在下).

Comment Style Requirements

Bilingual Format (Chinese first, English second):

# 要求：[作业要求中文]
# [技术解释中文]
# Requirement: [Assignment requirement in English]
# [Technical explanation in English]

Key Principles:

• •Start with # 要求： to mark assignment requirements (Chinese first)
• •Explain WHY technical choices are made
• •Provide parameter explanations
• •Mention alternative approaches when relevant
• •Chinese first, then English translation (中文在上，英文在下)
• •Keep comments concise but informative

Example:

# 要求：将图像从 RGB 转换为灰度图
# 从文件加载图像
# Requirement: Convert image from RGB to Grayscale
# Load image from file
img = cv2.imread(image_path)
gray_image = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# 要求：应用高斯模糊
# 边缘检测前降噪以防止图像噪声导致的误判边缘点
# 注意：高斯模糊使用加权平均，适合一般降噪
# 替代方案：中值模糊（cv2.medianBlur）使用中值，更适合椒盐噪声
# Requirement: Apply Gaussian blur
# Reduce noise before edge detection to prevent false edge points caused by image noise
# Note: Gaussian blur uses weighted average, good for general noise reduction
# Alternative: Median blur (cv2.medianBlur) uses median value, better for salt-and-pepper noise
blurred_image = cv2.GaussianBlur(img, (blur_ksize, blur_ksize), 0)

Docstring Format:

def function_name(param):
    """
    中文函数说明。
    English function description.
    
    中文详细解释：
    English detailed explanation:
    -------------------------------------------------------------------------
    中文内容
    English content
    -------------------------------------------------------------------------
    
    Args:
        param: 中文参数说明 / English parameter description
    
    Returns:
        中文返回值说明
        English return description
    """

For complete example: See Lab3 code at courses/mv/labs/lab3_orb.ipynb

Python File Format for Jupyter Conversion

When creating .py files that will be converted to Jupyter notebooks, use jupytext-compatible format:

File Structure:

# ---
# jupyter:
#   jupytext:
#     text_representation:
#       extension: .py
#       format_name: light
# ---

# # Lab Title - Main Heading
#
# **Objective:**
# Description of lab objectives
#
# **Materials Required:**
# - Item 1
# - Item 2
#
# **Lab Duration:** X hours

import cv2
import numpy as np
from matplotlib import pyplot as plt

# ## Part 1: Section Title
#
# Brief description of this part
#
# **Exercise 1:** Exercise description

def function_name(params):
    # Requirement: What this code does
    # Technical explanation
    # 要求：这段代码做什么
    # 技术解释
    code_here

# ## Test Exercise 1: Title
#
# Description of what this test does

# +
# 测试练习 1
# Test code for Exercise 1
print('Exercise 1: ...')
result = function_name(params)
# -

Key Rules:

• •Use # prefix for Markdown content (becomes Markdown cells)
• •Use # # for main headings, # ## for subheadings
• •Use # with blank line for paragraph breaks
• •Regular # comments inside functions (stays as code comments)
• •File header with jupytext metadata ensures proper conversion
• •Use # + and # - to mark cell boundaries for test code blocks

Cell Separation:

• •Each function definition: separate code cell
• •Each test/exercise execution: separate code cell with Markdown header
• •Use # + at start and # - at end to explicitly mark cell boundaries
• •Markdown sections (# ## ...) automatically create new cells

Conversion Command:

uv run jupytext --to notebook file.py -o file.ipynb

Why This Format:

• •✅ File header becomes Markdown cell (not code)
• •✅ Section titles become Markdown cells
• •✅ Exercise descriptions become Markdown cells
• •✅ Function code becomes code cells
• •✅ Bilingual comments preserved in code cells

Image Display Rule:

• •Always use plt.show() in test cells to display results directly in the Notebook.
• •Avoid using plt.close() in Notebook-targeted code as it prevents visual output.
• •Optional: Keep plt.savefig() if a permanent record is needed, but plt.show() must follow it or be used standalone.

For complete example: See courses/mv/labs/CST8508_Lab2.py

Core Workflows

Understanding Concepts

Ask for explanations at your level (beginner/intermediate/advanced). Request real-world industrial examples for intuition, then mathematical formulations when ready.

For detailed concept explanations: See references/concepts.md

Analyzing Code

Share OpenCV/Python code snippets and specify what you want to understand (algorithm flow, parameter tuning, optimization). Ask about common implementation pitfalls.

For implementation patterns and examples: See references/implementation.md

Completing Homework

Describe the vision problem and what you've tried. Ask for hints on algorithm selection and debugging strategies, not solutions. Verify your approach before implementing.

Running Experiments

1. •Define vision task goal
2. •Set up baseline algorithm
3. •Change one parameter at a time
4. •Analyze results with metrics and visualizations

For experiment templates: See references/experiments.md

Testing Knowledge

Specify topics (filtering, edge detection, segmentation, etc.), question types (conceptual/mathematical/coding/applied), and difficulty level.

Reviewing Material

Request summaries, algorithm comparison tables, parameter cheat sheets, or processing pipeline diagrams.

For quick reference: See references/quick-ref.md

Planning Projects

Follow phases: Requirements → Algorithm Selection → Implementation → Testing → Optimization

For project ideas and structure: See references/projects.md

Reading Papers

Use three-pass approach: (1) Abstract/figures/conclusion, (2) Problem/method/experiments, (3) Mathematical details/implementation

For key papers: See references/papers.md

Common Pitfalls

• •Poor lighting conditions not considered
• •Incorrect image preprocessing order
• •Wrong kernel size for filtering
• •Not handling edge cases in segmentation
• •Ignoring real-time performance constraints
• •Inadequate camera calibration

Best Practices

• •Start with simple algorithms before complex ones
• •Visualize intermediate processing steps
• •Test with diverse image conditions
• •Use established libraries (OpenCV, scikit-image)
• •Benchmark against baseline methods
• •Consider lighting and hardware constraints