Cocoa-Hazelnut Spread Jar Contaminant Detection via Microwave Sensing at 10.5 GHz

2400 instances of scattering parameters measured at 10.5 GHz to detect 10 types of contaminants in packaged food jars. The dataset, created by Luca Urbinati, is part of a series of five datasets each measured at a different microwave frequency. It contains 1200 uncontaminated and 1200 contaminated samples, with contaminants including metal, glass, and plastic objects placed at the surface or middle of the spread.

Use Cases

• Training binary classifiers for contaminant detection based on 30 scattering parameters.
• Developing multi-class models to identify specific contaminant types (e.g., glass_fragment_surface, small_metal_sphere_middle).
• Comparing model performance across different microwave frequencies using the related dataset series.
• Researching contactless sensing methods for packaged food inspection.

Strengths

• Contains 2400 total instances with a balanced 1200/1200 split between contaminated and uncontaminated samples.
• Includes 10 distinct contaminant classes with specific instance counts per class, enabling detailed multi-class analysis.
• Dataset is part of a controlled series measured at five specific frequencies (9.0, 9.5, 10.0, 10.5, 11.0 GHz), allowing for frequency-dependent study.

Limitations

• Column-level documentation is absent; field semantics for the 30 scattering parameters must be inferred after download.
• Last update date is unknown; freshness unverified.
• The description does not specify the physical setup or sensor calibration details, which may affect reproducibility.

Provenance

Source

Luca Urbinati via OpenML

Collection Method

Microwave sensing measurements of scattering parameters from a network of antennas, target object (jar), and medium.