%0 Journal Article
%A Torrecilla Velasco, José Santiago
%A Pérez Calabuig, Ana M.
%A Pradana Lopez, Sandra
%A Ramayo Muñoz, Andrea
%A Cancilla, John C.
%T Deep quantification of a refined adulterant blended into pure avocado oil
%D 2023
%U https://hdl.handle.net/20.500.14352/129753
%X This record corresponds to the peer-reviewed journal article “Deep quantification of a refined adulterant blended into pure avocado oil”, published in Food Chemistry (Volume 404, 2023), a Journal Citation Reports (JCR) indexed journal.The work presents the development and validation of an innovative, non-destructive methodology for the detection and quantification of avocado oil adulteration with refined olive oil by combining optical image acquisition and deep learning techniques. A comprehensive image database comprising 1,800 photographs of pure and adulterated samples (1–15% v/v) was generated under controlled lighting conditions using different shutter speeds to simulate real-world inspection scenarios.Convolutional neural networks based on transfer learning (ResNet34 architecture) were trained and optimized to perform both qualitative and quantitative classification tasks. The proposed models achieved high classification accuracies (~95%) during blind validation, demonstrating strong robustness, sensitivity, and quantitative discrimination capability, even at low adulterant concentrations.Compared to conventional analytical techniques, the proposed approach offers a cost-effective, rapid, and non-invasive alternative that does not require complex sample preparation or specialized laboratory equipment. The methodology shows strong potential for in-situ quality control, food fraud detection, and real-time monitoring across different stages of the production and distribution chain.
%~