Quentin Geissmann

Abstract Although visual features have been the cornerstone of insect recognition, morphological traits are often overlooked by modern "scale-invariant" deep learning-based methods. One such trait is absolute body size, which is commonly used by entomologists to describe species. We propose a novel approach that integrates explicit size information into computer vision models.

Insect Tuboid classifier To classify multiframe insect representations (i.e., tuboids), we based the Insect Tuboid Classifier (Fig 3), on a Residual Neural Network (ResNet) architecture [18] with 2 important modifications: (i) We explicitly included the size of the putative insect as an input variable to the fully connected layer as size may be important for classification and our images have consistent scale; and (ii) Since tuboid frames provide nonredundant information for classification...