Loading...
Research Project
Center for Electronics, Optoelectronics and Telecommunications
Funder
Authors
Publications
A tutorial on automatic hyperparameter tuning of deep spectral modelling for regression and classification tasks
Publication . Passos, Dário; Mishra, Puneet
Deep spectral modelling for regression and classification is gaining popularity in the chemometrics domain. A major topic in the deep learning (DL) modelling of spectral data is the choice and optimization of the deep neural network architecture suitable for the specific task of spectral modelling. Although there are several recent research articles already available in the chemometric domain showing advanced approaches to deep spectral modelling, currently, there is a lack of hands-on tutorial articles in this space that supply the non-expert user with practical tools to learn and implement advanced DL optimization methodologies aimed a spectral data. Hence, this tutorial article aims a reducing the gap between the non-expert user of DL in the chemometric community and the implementation of DL models for daily usage. This tutorial supplies a quick introduction to the state-of-the-art deep spectral modelling and related DL concepts and presents a set of methodologies aimed a DL hyperparameters' optimization. To this end, this tutorial shows two practical examples on how to implement and optimize two DL models for spectral regression and classification tasks. The models are implemented in python and Tensorflow and the complete code is supplied in the form of two complementary notebooks.
SpectraNet–53: A deep residual learning architecture for predicting soluble solids content with VIS–NIR spectroscopy
Publication . A. Martins, J.; Guerra, Rui Manuel Farinha das Neves; Pires, R.; Antunes, M.D.; Panagopoulos, T.; Brázio, A.; Afonso, A.M.; Silva, L.; Lucas, M.R.; Cavaco, A.M.
This work presents a new deep learning architecture, SpectraNet-53, for quantitative analysis of fruit spectra, optimized for predicting Soluble Solids Content (SSC, in Brix). The novelty of this approach resides in being an architecture trainable on a very small dataset, while keeping a performance level on-par or above Partial Least Squares (PLS), a time-proven machine learning method in the field of spectroscopy. SpectraNet-53 performance is assessed by determining the SSC of 616 Citrus sinensi L. Osbeck 'Newhall' oranges, from two Algarve (Portugal) orchards, spanning two consecutive years, and under different edaphoclimatic conditions. This dataset consists of short-wave near-infrared spectroscopic (SW-NIRS) data, and was acquired with a portable spectrometer, in the visible to near infrared region, on-tree and without temperature equalization. SpectraNet-53 results are compared to a similar state-of-the-art architecture, DeepSpectra, as well as PLS, and thoroughly assessed on 15 internal validation sets (where the training and test data were sampled from the same orchard or year) and on 28 external validation sets (training/test data sampled from different orchards/years). SpectraNet-53 was able to achieve better performance than DeepSpectra and PLS in several metrics, and is especially robust to training overfit. For external validation results, on average, SpectraNet-53 was 3.1% better than PLS on RMSEP (1.16 vs. 1.20 Brix), 11.6% better in SDR (1.22 vs. 1.10), and 28.0% better in R2 (0.40 vs. 0.31).
Sensory evaluation and spectra evolution of two kiwifruit cultivars during cold storage
Publication . Afonso, Andreia M.; Guerra, Rui; Cruz, Sandra; Antunes, Maria D.
Kiwifruit consumption has increased due to its rich nutritional properties. Although ‘Hayward’ continues to be the main cultivar, others, such as yellow fleshed ‘Jintao’, are of increasing interest. The objective of this research was to evaluate the acceptability and storage performance of these two cultivars. Sensory evaluation of green ‘Hayward’ and yellow ‘Jintao’ kiwifruit were performed along cold storage for three seasons/years to follow the organoleptic characteristics through ripening, as well as the acquisition of their spectra by Vis-NIR. For ‘Jintao’ were performed two sensory evaluations per year at 2.5- and 4.5-months’ storage and for ‘Hayward’ at 2.5-, 4.5- and 5.5-months’ storage. The nonparametric Mann–Whitney test and Kruskal–Wallis ANOVA were performed to test the significant differences between the mean ranks among the storage time. A non-metric multidimensional scaling plot method using the ALSCAL algorithm in a seven-point Likert scale was applied to determine the relationships in the data, and a new approach using the receiver operating characteristic (ROC) analysis was tested. The last revealed that, for both cultivars, sweetness, acidity and texture were the variables with better scores for General flavor. Aroma was also important on ‘Jintao’. A strong correlation between soluble solids content (SSC) and reflectance was found for both cultivars, with the 635–780 nm range being the most important. Regarding firmness, a good correlation with reflectance spectra was observed, particularly in ‘Hayward’ kiwifruit. Based on these results, Vis-NIR can be an objective alternative to explore for determination of the optimum eating-ripe stage.
Editorial: recent advancements on the development and ripening of Mediterranean fruits and tree crops
Publication . El-kereamy, Ashraf; Caruso, Marco; Torres, Carolina A.; Ana, Cavaco
The Mediterranean basin and other Mediterranean-type ecosystems (MTE) are home to many tree crops domesticated and adapted well to their environment. Several of them present specific development and ripening traits that challenge established models. Climate changes that are occurring in the Mediterranean area and in other MTE tends to aggravate the already irregular rainfall and temperature patterns, posing detrimental outcomes on crop performance, productivity, and changes in fruit ripening. With these climate changes, one would expect changes in the fruits and tree crops components growing in these ecosystems. Currently, we are experiencing a tremendous advance in the technology that allows researchers to study in-depth the basic phenomenon and find significant novel data to establish guidelines for new cultural practices, breeding programs, and variety selection that can better adapt to the changing conditions. The goal of this Research Topic was to highlight recent studies on the anatomical, physiological, metabolomic, and genomic processes occurring throughout the development and ripening of fruits and tree crops grown in the Mediterranean Basin and MTE, from field until postharvest. Since many of them are perennial species, they are subjected to adverse environmental conditions throughout their entire life cycle. Thus, the effect of cultural practices, varying environmental factors, as well as the impact of the various stresses on the performance of these tree crops were also acknowledged.
Estimation of soluble solids content and fruit temperature in 'rocha' pear using Vis-NIR spectroscopy and the spectraNet–32 deep learning architecture
Publication . Martins, J. A; Rodrigues, Daniela; Cavaco, A. M.; Antunes, Maria Dulce; Guerra, Rui Manuel Farinha das Neves
Spectra-based methods are becoming increasingly important in Precision Agriculture as they offer non-destructive, quick tools for measuring the quality of produce. This study introduces a novel approach for esti-mating the soluble solids content (SSC) of 'Rocha' pears using the SpectraNet-32 deep learning architecture, which operates on 1D fruit spectra in the visible to near-infrared region (Vis-NIRS). This method was also able to estimate fruit temperatures, which improved the SSC prediction performance. The dataset consisted of 3300 spectra from 1650 'Rocha' pears collected from local markets over several weeks during the 2010 and 2011 seasons, which had varying edaphoclimatic conditions. Two types of partial least squares (PLS) feature selection methods, under various configurations, were applied to the input spectra to identify the most significant wavelengths for training SpectraNet-32. The model's robustness was also compared to a similar state-of-the-art deep learning architecture, DeepSpectra, as well as four other classical machine learning algorithms: PLS, multiple linear regression (MLR), support vector machine (SVM), and multi-layer perceptron (MLP). In total, 23 different experimental method configurations were assessed, with 150 neural networks each. SpectraNet-32 consistently outperformed other methods in several metrics. On average, it was 6.1% better than PLS in terms of the root mean square error of prediction (RMSEP, 1.08 vs. 1.15%), 7.7% better in prediction gain (PG, 1.67 vs. 1.55), 3.6% better in the coefficient of determination (R2, 0.58 vs. 0.56) and 5.8% better in the coefficient of variation (CV%, 8.35 vs. 8.86).
Organizational Units
Description
Keywords
Contributors
Funders
Funding agency
Fundação para a Ciência e a Tecnologia
Funding programme
6817 - DCRRNI ID
Funding Award Number
UIDP/00631/2020