Smarter Plot Grading in CodeGrade: Two Automated Approaches for Faster Feedback

Method 2: Object-Based Plot Inspection

"Did they plot the right data in the right way?"

This method goes deeper. Instead of comparing images, it inspects the actual matplotlib objects created by the student, checking for things like titles, labels, plotted data, and legends.

Best of all: it all runs inline using CodeGrade’s Run in Notebook block — no external files required.

How it works:

A key part of this setup is a custom monkeypatching script that temporarily overrides matplotlib.pyplot.figure and plt.show. This lets you intercept and inspect all plots programmatically.

Here’s what the patch does:

• Captures every figure created, so they can be accessed after plotting
  
  
• Prevents plt.show() from blocking the test (since it runs in a headless environment)
  
  
• Clears figures after showing, just like in a regular plotting session

This makes it possible to write tests that assert:

• If a title, x/y labels, and legend exist
  
  
• If the correct number of lines or data points are plotted
  
  
• If the plotted values match an expected result (using NumPy comparison)

It’s precise, reusable, and gives you full control, making it perfect for assignments focused on interpreting or validating data.

Best for: verifying structure, data correctness, scientific plotting
Keep in mind: it requires matplotlib, and a little upfront scripting (which you can reuse across assignments).

Okay, I understand. Here's the updated text with the bullet points no longer in bold, making it look cleaner and less visually heavy:

Which Method Should You Use?

Here's a guide to help you choose the right method based on your needs:

If you care about...

• Matching appearance closely:
  • Use the Image Comparison method.
• Verifying structure and data correctness:
  • Use the Object Inspection method.
• Avoiding file management:
  • Use the Object Inspection method.
• Fast visual checks at scale:
  • Use the Image Comparison method.
• Detailed feedback on plotting conventions:
  • Use the Object Inspection method.

Combining Both Methods:You can also combine both methods for comprehensive testing:

• Use Object Inspection to validate correctness and structure.
• Use Image Comparison to catch visual styling issues.

Manual Plot Grading

If students use non-standard plotting libraries like Plotly, Bokeh, or create highly customized plots, automated grading might not cover everything.

CodeGrade makes it easy to grade these manually:

• Students save plots using plt.savefig(...)
  
  
• Use Output block to show saved plots in the Autotest output panel, right alongside their code and answers
  

• You assign points using a rubric and leave feedback as needed
  
  

This gives you the flexibility to still assess creative visualizations while maintaining consistency and clarity in grading.

‍

Plot grading doesn’t have to be slow or vague. Start with one method, combine both, or adapt them to your own course, and make grading smarter, faster, and more accurate.