Shazam's song recognition is proprietary, with no transparency on how it works. No open source models or benchmarks to compare against. This is bad science and software engineering - hiding algorithms behind a paywall. Shazam should release models for peer review and open source inference code.
