The data and findings discussed here come from the cited paper and do not represent test results from VISPEK equipment.
Recycled plastics are identified not only by material category, but also by cleanliness, color and transparency. This study split PET judgment into multiple tasks and compared the combination of near-infrared features and neural networks. Faced with higher classification results, it is still necessary to distinguish between the scale of data records and the scale of independent physical samples.
Summary
The study used near-infrared spectroscopy to train PET cleaning status, material category and transparency classifiers respectively, and compared various network models. Each task reports a high accuracy, but the number of records is not equal to the number of independent plastic bottles. The sample and data segmentation caliber is not yet clear. It is necessary to prevent repeated spectra of the same object from spanning training and testing.
(AI-generated conceptual illustration for context; it is not data from the paper, a photograph of the research equipment, or a VISPEK product diagram.)
Why It Matters
Splitting complex sorting goals into specific judgment tasks helps to set acceptance criteria for material identification and appearance status respectively.
Methods and Reference Measurements
The original article uses DLP NIRscan Nano to record 900–1700 nm data, and describes 295327 records containing fields such as wavelength, absorbance, and category. Spectral features were extracted after SG filtering and baseline processing, and three classifiers of PET clean state, PET and other materials, and colored and transparent PET were trained with HAttFFNN to compare various deep learning models and confusion matrices.
Wavelengths and Acquisition Parameters
| Band/parameter | Original meaning and usage instructions |
|---|---|
| 900–1700 nm | The original text collection or actual modeling scope of this case |
Key Findings
The accuracy rates of the three stages are 99.33%, 99.32% and 99.28% respectively. The number of data records and the way they are segmented are not equivalent to the scale of external verification of independently sourced plastic bottles.
Limitations and Interpretation
The number of independent physical samples is unclear, and it is necessary to focus on preventing repeated spectra of the same plastic bottles from being scattered in the training and test sets.
Band and model boundaries: This example uses original measurements or independent verification branches that lie entirely within 250–1700 nm.
Image copyright belongs to the paper’s original authors. License: CC BY 4.0.
Source Paper
Alimbekova N, Rai HM, Turymbetov T, Zhumadillayeva A. (2025). HAttFFNN: Hybridized attention mechanism-based feedforward neural network deep learning model for the plastic material classification of three stage materials on spectroscopic data. PLOS One 20(12). https://doi.org/10.1371/journal.pone.0336927
Contact Us
Whether portable equipment can be used for similar detection needs to be evaluated based on samples and experimental conditions, and the specific effect still needs to be verified through actual testing. Welcome to contact VISPEK to exchange your sample and testing needs, jointly evaluate experimental plans, and explore the feasibility of using our company's equipment for actual measurement and verification.
- Email: contact@vispek.cn
- Website: https://www.vispek.cn
VISPEK products shown for illustration; these were not the devices used in the cited study.