Fish quality is primarily effected by the number of days elapsed since harvesting, while bad storage conditions can also lead to quality degradation similar to the impact time. Existing approaches require laboratory testing, a laborious and timeconsuming process. In this work, we investigate technologies for quantifying fish quality though the development of deep learning models for analyzing imagery of fish. We first demonstrate that such a quantification is possible, to a certain degree, from multispectral images provided a sufficient number of training examples is available. Given that, we explore how knowledge distillation can be utilized for achieving similar fish quality estimation accuracy, but instead of using high-end multispectral imaging systems, using off-the-shelf RGB cameras. Experimental evaluation on individuals from the Mullus Marbatus family demonstrates that the proposed methodology constitutes a valid approach.
Quantification of food quality is a critical process for ensuring public health. Fish correspond to a particularly challenging case due to its high perishable nature as food. Existing approaches require laboratory testing, a laborious and timeconsuming process. In this paper, we propose a novel approach for evaluating fish freshness by exploiting the information encoded in the spectral profile acquired by a snapshot spectral camera. To extract the relevant information, we employ state-ofthe- art Convolutional Neural Networks and treat the problem as an instance of multi-class classification, where each class corresponds to a two-day period since harvesting. Experimental evaluation on individuals from the Sparidae (Boops sp.) family demonstrates that the proposed approach constitutes a valid methodology, offering both accuracy as well as effortless application.