Roman-David Kulko
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Roman-David Kulko has a total of 5 co-authors. Here are the top co-authors by number of shared publications:
- Benedikt Elser
Technologie Campus Grafenau, Technische Hochschule Deggendorf, 94481 Grafenau, Germany
5 shared publications - Alexander Pletl
Technologie Campus Grafenau, Technische Hochschule Deggendorf, 94481 Grafenau, Germany
4 shared publications - Andreas Hanus
Sesotec GmbH, Regener Straße 130, 94513 Schönberg, Germany
4 shared publications - Florian Wahl
Deggendorf Institute of Technology
1 shared publication - Heike Mempel
Institut für Gartenbau, Hochschule Weihenstephan-Triesdorf, 85354 Freising, Germany
1 shared publication
Latest Publications
- Open Access
Data-Efficient Polymer Classification Using Spectra Simulation and Bayesian Optimization
Plastic recycling represents an essential element of strategies aimed at lowering global carbon emissions while supporting a circular plastics economy. However, the effectiveness of current plastic sorting systems remains limited by data scarcity, spectral variability, and the complexity of real world waste streams. This study introduces a CNN-based polymer classification framework that integrates physics-informed spectral simulation, adaptive data augmentation, and Bayesian hyperparameter optimization to enable robust classification under data limited conditions. Our framework combines near-infrared (NIR) spectral data from technical scale measurements with synthetically generated spectra. With only 100 measured spectra per polymer, the proposed framework achieves average balanced accuracies of 0.9739 in multi-target polymer classification tasks. By using technical scale spectral data, this study bridges the gap between laboratory model development and real sorting conditions.
Open AccessData-Efficient Polymer Classification Using Spectra Simulation and Bayesian Optimization
Plastic recycling represents an essential element of strategies aimed at lowering global carbon emissions while supporting a circular plastics economy. However, the effectiveness of current plastic sorting systems remains limited by data scarcity, spectral variability, and the complexity of real world waste streams. This study introduces a CNN-based polymer classification framework that integrates physics-informed spectral simulation, adaptive data augmentation, and Bayesian hyperparameter optimization to enable robust classification under data limited conditions. Our framework combines near-infrared (NIR) spectral data from technical scale measurements with synthetically generated spectra. With only 100 measured spectra per polymer, the proposed framework achieves average balanced accuracies of 0.9739 in multi-target polymer classification tasks. By using technical scale spectral data, this study bridges the gap between laboratory model development and real sorting conditions.
- Open Access
Generative Artificial Intelligence for Synthetic Spectral Data Augmentation in Sensor-Based Plastic Recycling
The reliance on deep learning models for sensor-based material classification amplifies the demand for labeled training data. However, acquiring large-scale, annotated spectral data for applications such as near-infrared (NIR) reflectance spectroscopy in plastic sorting remains a significant challenge due to high acquisition costs and environmental variability. This paper investigates the potential of large language models (LLMs) in synthetic spectral data generation. Specifically, it examines whether LLMs have acquired sufficient implicit knowledge to assist in generating spectral data and introduce meaningful variations that enhance model performance when used for data augmentation. Classification accuracy is reported exclusively as a proxy for structural plausibility of the augmented spectra; maximizing augmentation performance itself is not the study’s goal. From as little as one empirical mean spectrum per class, LLM-guided simulation produced data that enabled up to 86% accuracy, evidence that the generated variation preserves class-distinguishing information. While the approach performs best for spectral distinct polymers, overlapping classes remain challenging. Additionally, the transfer of optimized augmentation parameters to unseen classes indicates potential for generalization across material types. While plastic sorting serves as a case study, the methodology may be applicable to other domains such as agriculture or food quality assessment, where spectral data are limited. The study outlines a novel path toward scalable, AI-supported data augmentation in spectroscopy-based classification systems.
Open AccessGenerative Artificial Intelligence for Synthetic Spectral Data Augmentation in Sensor-Based Plastic Recycling
The reliance on deep learning models for sensor-based material classification amplifies the demand for labeled training data. However, acquiring large-scale, annotated spectral data for applications such as near-infrared (NIR) reflectance spectroscopy in plastic sorting remains a significant challenge due to high acquisition costs and environmental variability. This paper investigates the potential of large language models (LLMs) in synthetic spectral data generation. Specifically, it examines whether LLMs have acquired sufficient implicit knowledge to assist in generating spectral data and introduce meaningful variations that enhance model performance when used for data augmentation. Classification accuracy is reported exclusively as a proxy for structural plausibility of the augmented spectra; maximizing augmentation performance itself is not the study’s goal. From as little as one empirical mean spectrum per class, LLM-guided simulation produced data that enabled up to 86% accuracy, evidence that the generated variation preserves class-distinguishing information. While the approach performs best for spectral distinct polymers, overlapping classes remain challenging. Additionally, the transfer of optimized augmentation parameters to unseen classes indicates potential for generalization across material types. While plastic sorting serves as a case study, the methodology may be applicable to other domains such as agriculture or food quality assessment, where spectral data are limited. The study outlines a novel path toward scalable, AI-supported data augmentation in spectroscopy-based classification systems.
- Open Access
A Plastic Classification Model Based on Simulated Data
Plastic recycling holds significant potential to reduce global carbon emissions. Despite advances in recycling technologies, challenges such as limited data availability, contamination in sorted materials, and the complexity of real-world material flows continue to hinder progress. This study addresses these issues by introducing a novel approach to plastic classification, leveraging simulated spectral data to reduce reliance on large datasets and improve classification accuracy. Using near-infrared spectroscopy and deep learning models, the framework integrates data augmentation techniques and spectral simulation to augment datasets with synthetic spectra based on a data sample of 25 plastic granules. The proposed classification framework achieves excellent recall and robust balanced accuracy for both binary and multi-target polymer classification with minimal data input (only 50 spectra per class). Thus, the measurement effort is drastically reduced while maintaining an equally high model accuracy. The model significantly outperforms conventional unsupervised approaches. By overcoming the limitations of supervised learning models, the proposed framework provides a scalable and efficient solution for plastics recycling.
Open AccessA Plastic Classification Model Based on Simulated Data
Plastic recycling holds significant potential to reduce global carbon emissions. Despite advances in recycling technologies, challenges such as limited data availability, contamination in sorted materials, and the complexity of real-world material flows continue to hinder progress. This study addresses these issues by introducing a novel approach to plastic classification, leveraging simulated spectral data to reduce reliance on large datasets and improve classification accuracy. Using near-infrared spectroscopy and deep learning models, the framework integrates data augmentation techniques and spectral simulation to augment datasets with synthetic spectra based on a data sample of 25 plastic granules. The proposed classification framework achieves excellent recall and robust balanced accuracy for both binary and multi-target polymer classification with minimal data input (only 50 spectra per class). Thus, the measurement effort is drastically reduced while maintaining an equally high model accuracy. The model significantly outperforms conventional unsupervised approaches. By overcoming the limitations of supervised learning models, the proposed framework provides a scalable and efficient solution for plastics recycling.
- Open Access
Detection of Plastic Granules and Their Mixtures
Chemically pure plastic granulate is used as the starting material in the production of plastic parts. Extrusion machines rely on purity, otherwise resources are lost, and waste is produced. To avoid losses, the machines need to analyze the raw material. Spectroscopy in the visible and near-infrared range and machine learning can be used as analyzers. We present an approach using two spectrometers with a spectral range of 400–1700 nm and a fusion model comprising classification, regression, and validation to detect 25 materials and proportions of their binary mixtures. one dimensional convolutional neural network is used for classification and partial least squares regression for the estimation of proportions. The classification is validated by reconstructing the sample spectrum using the component spectra in linear least squares fitting. To save time and effort, the fusion model is trained on semi-empirical spectral data. The component spectra are acquired empirically and the binary mixture spectra are computed as linear combinations. The fusion model achieves very a high accuracy on visible and near-infrared spectral data. Even in a smaller spectral range from 400–1100 nm, the accuracy is high. The visible and near-infrared spectroscopy and the presented fusion model can be used as a concept for building an analyzer. Inexpensive silicon sensor-based spectrometers can be used.
Open AccessDetection of Plastic Granules and Their Mixtures
Chemically pure plastic granulate is used as the starting material in the production of plastic parts. Extrusion machines rely on purity, otherwise resources are lost, and waste is produced. To avoid losses, the machines need to analyze the raw material. Spectroscopy in the visible and near-infrared range and machine learning can be used as analyzers. We present an approach using two spectrometers with a spectral range of 400–1700 nm and a fusion model comprising classification, regression, and validation to detect 25 materials and proportions of their binary mixtures. one dimensional convolutional neural network is used for classification and partial least squares regression for the estimation of proportions. The classification is validated by reconstructing the sample spectrum using the component spectra in linear least squares fitting. To save time and effort, the fusion model is trained on semi-empirical spectral data. The component spectra are acquired empirically and the binary mixture spectra are computed as linear combinations. The fusion model achieves very a high accuracy on visible and near-infrared spectral data. Even in a smaller spectral range from 400–1100 nm, the accuracy is high. The visible and near-infrared spectroscopy and the presented fusion model can be used as a concept for building an analyzer. Inexpensive silicon sensor-based spectrometers can be used.
- Open Access
OpenVNT: An Open Platform for VIS-NIR Technology
Spectrometers measure diffuse reflectance and create a “molecular fingerprint” of the material under investigation. Ruggedized, small scale devices for “in-field” use cases exist. Such devices might for example be used by companies in the food supply chain for inward inspection of goods. However, their application for the industrial Internet of Things workflows or scientific research is limited due to their proprietary nature. We propose an open platform for visible and near-infrared technology (OpenVNT), an open platform for capturing, transmitting, and analysing spectral measurements. It is built for use in the field, as it is battery-powered and transmits data wireless. To achieve high accuracy, the OpenVNT instrument contains two spectrometers covering a wavelength range of 400–1700 nm. We conducted a study on white grapes to compare the performance of the OpenVNT instrument against the Felix Instruments F750, an established commercial instrument. Using a refractometer as ground truth, we built and validated models to estimate the Brix value. As a quality measure, we used coefficient of determination of the cross-validation () between the instrument estimation and ground truth. With 0.94 for the OpenVNT and 0.97 for the F750, a comparable was achieved for both instruments. OpenVNT matches the performance of commercially available instruments at one tenth of the price. We provide an open bill of materials, building instructions, firmware, and analysis software to enable research and industrial IOT solutions without the limitations of walled garden platforms.Open AccessOpenVNT: An Open Platform for VIS-NIR Technology
Spectrometers measure diffuse reflectance and create a “molecular fingerprint” of the material under investigation. Ruggedized, small scale devices for “in-field” use cases exist. Such devices might for example be used by companies in the food supply chain for inward inspection of goods. However, their application for the industrial Internet of Things workflows or scientific research is limited due to their proprietary nature. We propose an open platform for visible and near-infrared technology (OpenVNT), an open platform for capturing, transmitting, and analysing spectral measurements. It is built for use in the field, as it is battery-powered and transmits data wireless. To achieve high accuracy, the OpenVNT instrument contains two spectrometers covering a wavelength range of 400–1700 nm. We conducted a study on white grapes to compare the performance of the OpenVNT instrument against the Felix Instruments F750, an established commercial instrument. Using a refractometer as ground truth, we built and validated models to estimate the Brix value. As a quality measure, we used coefficient of determination of the cross-validation () between the instrument estimation and ground truth. With 0.94 for the OpenVNT and 0.97 for the F750, a comparable was achieved for both instruments. OpenVNT matches the performance of commercially available instruments at one tenth of the price. We provide an open bill of materials, building instructions, firmware, and analysis software to enable research and industrial IOT solutions without the limitations of walled garden platforms.