Analysis and classification of footwear line drawings: research on fashionattributes using computer vision algorithms

With the rapid evolution of fashion trends and consumer preferences, the imperative for agility in footwear design has become increasingly pronounced. Central to the design process was the criticality of shoe line drawings, the burgeoning advancements in computer vision and deep learning technologies have engendered a wealth of research in fashion element recognition. Regrettably, the application of such advancements to footwear remains relatively underexplored. This study introduces a novel computer vision system tailored to discern and categorise footwear line drawings. The methodology entails the preliminary training of Mask R-CNN for shoe body extraction from footwear imagery, followed by applying the PIDINet edge detection algorithm for line drawing delineation, culminating in utilising a classification model for line drawing. Encouragingly, our findings evince the system’s adeptness in successful line drawing extraction and classification, particularly demonstrating heightened accuracy in differentiating distinct styles such as nude shoes, boots, and slippers characterized by salient outline features. This pioneering endeavour not only addresses a gap in footwear element recognition research but also circumvents the need for an extensive footwear database for algorithmic training. The anticipated automation of algorithmic footwear line drawing recognition holds promise for enhancing operational efficiency and innovation, fostering sustainable advancements in fashion research.