One-dimensional CNNs for Near-Infrared Prediction of Protein and Moisture in Cereal Grains: The Effects of Architecture and Input Preparation
Abstract
Near-infrared (NIR) spectroscopy is widely used for the rapid, non-destructive determination of constituents such as protein and moisture in cereal grains, but model comparisons in this field are often confounded by differences in the inputs received by each model. We benchmark a compact one-dimensional CNN derived by one-factor-at-a-time ablation (CNN-Baseline) and a randomly searched CNN (CNN-RS) against PLSR, SVR, XGBoost on six cereal-grain datasets (n=500-5046) under three input conditions: raw spectra, optimal preprocessing, and preprocessing plus wavelength selection. A single-filter with a kernel size of 11 of convolution and batch normalization proved sufficient. Given a unified raw input, CNN-RS was most accurate on five of the six datasets, while CNN-Baseline came within 0.003-0.017 in test R2 on the four larger datasets. PLSR led only on the smallest protein dataset (XDS, n=500) and was the most stable model. Preprocessing had almost no effect on CNN accuracy on the four larger datasets, but improved accuracy by 0.034-0.047 (CNN-Baseline) and 0.017-0.055 (CNN-RS) on the two smallest, and brought the networks to their optimum in a median of 49% fewer epochs. Therefore, replacing explicit preprocessing with learned filters was only justified when sufficient training data were available. Once each model received its own preprocessing and wavelength subset, SVR became most accurate on four of six datasets and XGBoost rose from the weakest model to within 0.01-0.03 of the best model on four of the six datasets, suggesting that the principal advantage of CNNs lied in robustness to raw input data rather than in attainable accuracy. Overall, model selection for NIR calibration of cereal grains depends more on sample size and the input preparation than on network architecture.
Related
- Perceptual repulsion and decisional attraction in male body serial dependence: a dimensional dissociation in channel architecture
- Rice multi-variety yield prediction using multi-temporal fusion of thermal infrared and multispectral UAV data based on random Forest feature selection
- Visible/Near infrared spectroscopy in muscle beef analysis: how device choice and sampling affect prediction accuracy
- Three-dimensional architecture of the endoplasmic reticulum network in soybean roots reveals extensive remodeling during cyst nematode parasitism
- Deciphering the dynamic genetic architecture of adult-plant salinity tolerance in bread wheat using multi-model GWAS and genomic prediction.
BibTeX
@misc{deng2026onedimensional,
title = {{One-dimensional CNNs for Near-Infrared Prediction of Protein and Moisture in Cereal Grains: The Effects of Architecture and Input Preparation}},
author = {G. Deng and W. Chi and P. Yu and F. WU},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.12.751160},
url = {https://www.biorxiv.org/content/10.64898/2026.09.12.751160v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - One-dimensional CNNs for Near-Infrared Prediction of Protein and Moisture in Cereal Grains: The Effects of Architecture and Input Preparation
AU - G. Deng
AU - W. Chi
AU - P. Yu
AU - F. WU
PY - 2026
DA - 2026/09/12
PB - bioRxiv
DO - 10.64898/2026.09.12.751160
UR - https://www.biorxiv.org/content/10.64898/2026.09.12.751160v1
AB - Near-infrared (NIR) spectroscopy is widely used for the rapid, non-destructive determination of constituents such as protein and moisture in cereal grains, but model comparisons in this field are often confounded by differences in the inputs received by each model. We benchmark a compact one-dimensional CNN derived by one-factor-at-a-time ablation (CNN-Baseline) and a randomly searched CNN (CNN-RS) against PLSR, SVR, XGBoost on six cereal-grain datasets (n=500-5046) under three input conditions: raw spectra, optimal preprocessing, and preprocessing plus wavelength selection. A single-filter with a kernel size of 11 of convolution and batch normalization proved sufficient. Given a unified raw input, CNN-RS was most accurate on five of the six datasets, while CNN-Baseline came within 0.003-0.017 in test R2 on the four larger datasets. PLSR led only on the smallest protein dataset (XDS, n=500) and was the most stable model. Preprocessing had almost no effect on CNN accuracy on the four larger datasets, but improved accuracy by 0.034-0.047 (CNN-Baseline) and 0.017-0.055 (CNN-RS) on the two smallest, and brought the networks to their optimum in a median of 49% fewer epochs. Therefore, replacing explicit preprocessing with learned filters was only justified when sufficient training data were available. Once each model received its own preprocessing and wavelength subset, SVR became most accurate on four of six datasets and XGBoost rose from the weakest model to within 0.01-0.03 of the best model on four of the six datasets, suggesting that the principal advantage of CNNs lied in robustness to raw input data rather than in attainable accuracy. Overall, model selection for NIR calibration of cereal grains depends more on sample size and the input preparation than on network architecture.
ER -