A convolutional neural network approach to early detect physiological disorders and maturity in ‘Abate Fétel’ pears

Image processing through the implementation of manually coded algorithms has been adopted to sort fruit depending on their defects and quality during postharvest operations. This study tested convolution neural networks with “You Only Look Once” (YOLO) architecture using a commercial online platform to detect physiological disorders and ripening stage in ‘Abbé Fétel’ pears. Storage disorders such as superficial scald and the starch pattern index (SPI) were assessed. Two different models were trained to detect: I) early symptoms of superficial scald or senescence scald on pear skin; II) the SPI value of pears assessed using the Lugol solution. Preliminary statistics showed that the first model reached low accuracy (up to 20% of true positives) but maintained a good average precision (60%) with different confidence thresholds (40 and 20%). The second had good performances compared to the CTIFL and Laimburg scales, with an F1 score of 0.36 and 0.59, respectively. The effectiveness of the transfer learning method was demonstrated. However, further image labeling and modeling research is needed to improve the performance of the simulations and to develop an application for portable devices for pre and postharvest quality assessment. These results could provide effective tools for producers to manage pears in different cold rooms and thus help them to ensure consumers acceptance.