The goal: test whether NIRS could detect meat composition through packaging in a real-world setting.
But I made a critical sampling mistake. The spectra came back noisy and inconsistent, and typical preprocessing methods like SNV, MSC, and Savitzky-Golay filtering weren’t enough to clean it up.
In this video, I break down:
• How poor sampling technique created unreliable data
• Why the instrument wasn’t the problem
• How I used Python and RMSE filtering to rescue the experiment and reveal meaningful patterns
🧠 Whether you’re into field-deployable spectroscopy, food analysis, or signal processing in Python—this is a real case study in how good tools still need good methods.
✅ Subscribe for more hands-on analytical science, Python tutorials, and real-world experimentation.