Global local fusion improves acne severity grading accuracy
CG-HAF: An Interpretable Global-Local Lesion-Burden Fusion Framework for Ordinal Acne Severity Grading in Agentic Skincare Support
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Grading how severe acne is can be hard because similar levels look very alike and need both a full-face view and attention to specific spots. The authors made a new method called CG-HAF that combines general face-level guesses with detailed counts and sizes of lesions to better decide the severity. This approach works better, especially for bad acne cases, and it also helps explain how the decision was made. They also found that grading systems differ across datasets, so matching criteria is important for using this method broadly.
What this means in practice
- •For skincare technology developers: Design software that grades acne severity by combining overall face patterns with counted lesion details for more accurate and explainable results.$Commercial implications: This method enables creation of sellable skincare apps providing clearer acne severity assessments to consumers and professionals.
- •For medical imaging developers: Create acne grading tools that separately assess global skin appearance and local lesion features to improve severity classification accuracy.
Authors
Muhammad Muhtasim Shahriar, Md. Naimur Asif Borno, Saad Aloteibi, Mohammad Ali Moni
Abstract
Ordinal acne severity grading requires distinguishing visually similar neighboring grades while jointly weighing holistic facial appearance and localized lesion burden - evidence that most existing approaches collapse into a single opaque representation. We introduce CG-HAF, a global-local fusion framework that instead keeps this evidence explicit: averaged holistic severity probabilities from independently trained classifiers are combined with structured lesion-burden descriptors from an object detector (lesion count, detection confidence, lesion area) into a compact representation, from which a lightweight, interpretable classifier produces the final grade. On a widely used benchmark, this fusion yields a clear, statistically supported improvement over global-evidence-only baselines, with the largest gains on the most severe cases. Testing on an independent dataset with a different grading standard shows that strong within-dataset performance does not transfer automatically, and a follow-up diagnostic attributes much of this gap to mismatched grading criteria rather than detection failure alone. These findings support interpretable global-local fusion as an effective strategy for ordinal acne grading while highlighting criterion alignment as key to cross-dataset portability, with a further illustration of how the resulting severity signal can support transparent, non-diagnostic decision-making in skincare applications.