Problem

The tests in tests/core/metrics/test_meitrcs_identity_accuracy.py were failing due to hard-coded identity_scores arrays that didn't match the number of individuals in the test cases. Specifically, one test case with 3 individuals had identity_scores arrays with only 2 values instead of 3, causing dimension mismatches.

Root Cause

The compute_identity_scores function expects identity_scores with shape (num_assemblies, num_bodyparts, num_individuals), but the test data had:

{
    "individuals": ["i1", "i2", "i3"],  # 3 individuals
    "identity_scores": {
        "img0.png": [
            [[0.7, 0.3], [0.6, 0.2]],  # Only 2 values, should be 3!
            # ...
        ],
    },
}

This created a shape mismatch: (3, 2, 2) instead of the expected (3, 2, 3).

Solution

Fixed the hard-coded test data by adding the missing third value to match the number of individuals:

{
    "individuals": ["i1", "i2", "i3"],  # 3 individuals  
    "identity_scores": {
        "img0.png": [
            [[0.7, 0.3, 0.0], [0.6, 0.2, 0.2]],  # Now has 3 values ✓
            [[0.6, 0.3, 0.1], [0.6, 0.2, 0.2]], 
            [[0.6, 0.3, 0.1], [0.6, 0.2, 0.2]], 
        ],
    },
}

Validation

• ✅ All 7 test cases in test_meitrcs_identity_accuracy.py now have correct dimensions
• ✅ Verified that 2-individual test cases remain correct with 2-element arrays
• ✅ Tested the fix works with 2, 3, and 4 individuals dynamically
• ✅ Confirmed test_apis_evaluate.py already uses proper dynamic data generation

Impact

• Tests should now pass without dimension-related failures
• No functional logic changed, only test data corrected
• Maintains backward compatibility for all existing functionality
• Enables proper testing of identity accuracy across different numbers of individuals

💬 Share your feedback on Copilot coding agent for the chance to win a $200 gift card! Click here to start the survey.