“…With increasing computational power, more complex neural network architectures, i.e., deep learning (DL) approaches have recently helped in tackling more challenging tasks in the field of food and agricultural science (Lee et al, 2015;DeChant et al, 2017;Lu et al, 2017;Zhang et al, 2018). Although there have been relatively fewer DL studies to identify filth elements for food contamination (Reinholds et al, 2015;Bansal et al, 2017), variations of DL designs such as Region-based Fully Convolutional Network (R-FCN), convolutional block attention module (CBAM), convolutional neural network (CNN) and pre-trained models have shown promising performances for pest, stored-grain insect, and fly classification (Chen et al, 2020;Kuzuhara et al, 2020;Shi et al, 2020). The DL models have not only achieved high classification accuracies, but also offered a new way of feature extraction embedded in the process as an alternative to conventional features such as domain-dependent, global, local, and mid-level features (Martineau et al, 2017).…”