Issue #75: Add MRI robustness benchmarking reports and documentation

[prerendrarahitya1708]

Issue #75

Summary

Implemented robustness benchmarking for the MRI tumor detection system.

Changes Made

• Added robustness benchmark evaluation workflow
• Generated confidence degradation reports
• Added corruption impact visualization
• Added model comparison documentation
• Updated robustness README with benchmark results

Corruptions Evaluated

• Gaussian Noise
• Blur
• Brightness Shift
• Low Resolution
• Compression Artifacts

Key Findings

• Gaussian noise produced the largest confidence degradation.
• Brightness changes had minimal impact.
• The model remained relatively stable under blur and compression.
• The benchmarking framework provides a baseline for future model robustness comparisons.

Outputs Generated

• robustness_report.json
• robustness_plot.png
• model_comparison.md

[Brijeshthummar02]

@prerendrarahitya1708 we need the same benchmarking for entire dataset and all MRI scans with all cells of notebook report.

[prerendrarahitya1708]

Updated the robustness benchmarking workflow to evaluate the entire MRI dataset (3954 scans) instead of a single MRI image.

Changes:

• Evaluated all MRI scans in the dataset.
• Applied Gaussian Noise, Blur, Brightness Shift, Low Resolution, and Compression corruptions to every scan.
• Generated updated robustness reports from dataset-wide evaluation.
• Updated README and model comparison documentation to reflect full-dataset benchmarking results.

This addresses the feedback requesting benchmarking across the entire dataset and all MRI scans.

[Brijeshthummar02]

resolve conflicts

[prerendrarahitya1708]

resolved conflicts

[Brijeshthummar02]

@prerendrarahitya1708 what was main benchmark spike you show after updates?

[Brijeshthummar02]

make sure entire set is updated after running all the cells of notebook and testing datasets.