Acoustic scattering AI classifies hair with 90% accuracy—then retracted

The study, originally submitted to arXiv in June 2025, introduced a novel method for non-invasive object classification using acoustic scattering. The researchers used a self-supervised deep learning approach, fine-tuning all parameters of a model to classify hair type and moisture from reflected sound signals. They benchmarked multiple strategies, including fully supervised learning and foundation model fine-tuning, with the self-supervised variant hitting nearly 90% accuracy. This technique offers a privacy-preserving, non-contact alternative to visual classification, with potential applications in cosmetics, hair care, and medical diagnostics.

However, the paper has been withdrawn by the authors (version v2, May 2026) due to a miscommunication that led to incomplete authorship and missing early contributions. The retraction highlights the importance of proper attribution in academic research. While the technical approach remains promising—acoustic scattering could replace cameras in sensitive settings—the withdrawal casts a shadow on the immediate credibility of the work. The underlying methodology, if replicated transparently, could still have significant industry impact for non-invasive material classification.