Evaluation and prediction of salt effects on pig muscle by deep UV and machine learning.

The salting process for meat transformation is a crucial step in conventional industry. Recent developments in label-free spectrometry techniques combined with machine learning hold great promise for high-precision salt processing. In this study, we applied UV fluorescence to characterize salting treatments in pig's Teres major muscle and predict NaCl concentrations. t-SNE analyses based on spectral measurements revealed clear differences between NaCl-free and salted treatments. However, salt treatments were not clearly identified. We then highlighted and exploited a variability seen in the emission spectra at the wavelengths 300, 318, and 360 nm, which reflected structural or compositional changes. Using this information, predictive models could accurately identify the five salted treatments with a high specificity and sensitivity or predict salt concentrations. This study paves the way toward the possibility for industrials to precisely adjust NaCl concentrations with precision during processing.

Influence of sodium chloride on muscle UV autofluorescence characteristics.

Salted and tumbled pork teres major muscle samples, with varying sodium chloride content (1.1 % to 1.9 %), were examined by UV fluorescence spectroscopy. Results indicated that muscle fluorescence varies with salt level as a consequence of the protein denaturation state. The 1.5 % NaCl level was the threshold beyond which the fluorescence properties no longer changed markedly. Changes in muscle fluorescence do not appear to be linearly related to salt levels. Hence, we explored whether the change in fluorescence relies on other factors relating to the variability of carcass characteristics and on muscle physicochemical changes that are partly dependent on stress response and on postmortem metabolism evolution.

Continual Learning Strategy in One-Stage Object Detection Framework Based on Experience Replay for Autonomous Driving Vehicle.

Object detection is an important aspect for autonomous driving vehicles (ADV), which may comprise of a machine learning model that detects a range of classes. As the deployment of ADV widens globally, the variety of objects to be detected may increase beyond the designated range of classes. Continual learning for object detection essentially ensure a robust adaptation of a model to detect additional classes on the fly. This study proposes a novel continual learning method for object detection that learns new object class(es) along with cumulative memory of classes from prior learning rounds to avoid any catastrophic forgetting. The results of PASCAL VOC 2007 have suggested that the proposed ER method obtains 4.3% of mAP drop compared against the all-classes learning, which is the lowest amongst other prior arts.