RingFIR: A Large Volume Earring Dataset for Fashion Image Retrieval

Abstract

Fashion image retrieval (FIR) is a challenging task, which involves similar item searching from a massive collection of fashion products based on a query image. FIR in different garments and shoes are popular in literature. More complex fashion products such as ornaments are getting less attention. Here, we introduce a new earring dataset, namely, RingFIR. The dataset is a collection of (∼2.6K) high-quality images collected from major India based jewellery chains. The dataset is labelled in 46 classes in a structured manner. We have benchmarked the dataset using state-of-the-art image retrieval methods. We believe that the dataset is challenging and will attract computer vision researchers in the future. The dataset is available publicly (https://github.com/ skarifahmed/RingFIR).