Search

Over the last years, deep learning (DL) models have led to an enormous breakthrough in a wide range of computer vision tasks, including the classification and quantification of internal defects in fruit using X-ray CT images. However, developing these models typically requires large annotated datasets. Obtaining such datasets faces multiple challenges as the availability of defect fruit is unpredictable and it is unclear whether the disordered fruit in the dataset are representative of all possible defects. This work proposes a method to circumvent these problems by generating synthetic CT images using a conditional Generative Adversarial Network (cGAN) that learns the underlying distribution of gray values for different classes ("healthy tissue", "internal browning", "background, core and cavities") in manually annotated images of pear fruit. The performance of the predictor was evaluated by quantitative metrics and visual inspection. To generate new synthetic data, operations are suggested and tested on annotated data to create as much variation as possible in the generated pears. The Fréchet Inception Distance (FID) and visual observations showed that the cGAN was effective in image-to-image translation, generating CT images of both healthy and disordered fruit based on their annotations. The best performing model obtained a low FID score of 10.2 and produced detailed visual features in the synthetic CT data, such as vascular bundles. The high potential and broad applicability was further demonstrated on augmented image annotations for generating new healthy and disordered pears with different appearances of defects. The proposed method can be used to produce additional training data when datasets are limited and laborious to manually annotate. Furthermore, the robustness of other DL models might increase by including new synthetic samples with a high variety of disorders during training. • A conditional GAN (pix2pix) was trained on X-ray CT data of healthy and defect pears and conditioned on its semantics. • The network is able to generate synthetic CT data of healthy and disorder pears with different levels of severity. • The synthetic data can be used as additional training data for robust classification and segmentation models. [ABSTRACT FROM AUTHOR]

During respiratory quotient based dynamic controlled atmosphere (RQ-DCA) storage, the oxygen concentration in the room is reduced until the RQ measured in the air of the room (R Q a i r ) increases above a predefined threshold. This increase in RQ indicates an increased extent of fermentation within the fruit. Because of internal gas gradients, the oxygen concentration inside the fruit is lower than that of the storage atmosphere and, hence, the RQ value in the fruit (R Q f r u i t , m a x ) is larger than R Q a i r . Moreover, due to biological variability and fluctuations in the storage conditions, both R Q f r u i t , m a x and R Q a i r vary. Currently, no information exists about how the RQ measured in the atmosphere correlates with the RQ value inside the fruit, and how representative this R Q a i r is for all fruit inside the room. In this article, we investigated how variability affects the critical R Q a i r threshold in RQ-DCA. Hereto we used a multiscale stochastic reaction-diffusion model to compute gas transport within pears of three different cultivars, 'Cepuna', 'Conference' and 'Queen's Forelle'. Biological variability in diffusion and respiration parameters and fluctuations in cool room conditions were taken into account by performing Monte-Carlo simulations. The results showed large variations in both R Q f r u i t , m a x and R Q a i r . The effect of cool room filling degree on both RQ values was negligible, while variations in fruit material properties (pear size, tissue gas diffusivity and respiration kinetics) showed to be of high importance. However, as R Q a i r increased before fermentation started to dominate, the underestimation and variability of R Q a i r did not inhibit a proper detection of hypoxia inside the fruit within the room. This finding indicates that R Q a i r is a proper parameter to control DCA storage of various cultivars. Due to differences in fruit properties, the RQ-response varied between the cultivars. 'Cepuna' requires the highest O 2 (0.61 kPa) to prevent fermentation, while 'Conference' and 'Queen's Forelle' can be stored at lower O 2 (0.33 and 0.22 kPa, respectively). The results obtained in this research can be used to optimize RQ-DCA storage of each cultivar. • Gas exchange properties of pear vary with cultivar. • Stochastic simulations quantified the variability in RQ during RQ-DCA storage. • Large variations in RQ were predicted during RQ-DCA storage. • Room RQ measurements allowed to properly detect fruit fermentation. [ABSTRACT FROM AUTHOR]