Deep Learning with Discriminative Margin Loss for Cross-Domain Consumer-to-Shop Clothes Retrieval

• Published in: Sensors
• Main Authors: Alirezazadeh, Pendar, Dornaika, Fadi, Moujahid, Abdelmalik
• Format: Article in Journal/Newspaper
• Language: English
• Published: MDPI 2022
• Subjects: cross-domain fashion retrieval; margin-based loss function; adaptive margin; deep learning; discriminative analysis; DML
• Online Access: http://hdl.handle.net/10810/56381; https://doi.org/10.3390/s22072660

Consumer-to-shop clothes retrieval refers to the problem of matching photos taken by customers with their counterparts in the shop. Due to some problems, such as a large number of clothing categories, different appearances of clothing items due to different camera angles and shooting conditions, different background environments, and different body postures, the retrieval accuracy of traditional consumer-to-shop models is always low. With advances in convolutional neural networks (CNNs), the accuracy of garment retrieval has been significantly improved. Most approaches addressing this problem use single CNNs in conjunction with a softmax loss function to extract discriminative features. In the fashion domain, negative pairs can have small or large visual differences that make it difficult to minimize intraclass variance and maximize interclass variance with softmax. Margin-based softmax losses such as Additive Margin-Softmax (aka CosFace) improve the discriminative power of the original softmax loss, but since they consider the same margin for the positive and negative pairs, they are not suitable for cross-domain fashion search. In this work, we introduce the cross-domain discriminative margin loss (DML) to deal with the large variability of negative pairs in fashion. DML learns two different margins for positive and negative pairs such that the negative margin is larger than the positive margin, which provides stronger intraclass reduction for negative pairs. The experiments conducted on publicly available fashion datasets DARN and two benchmarks of the DeepFashion dataset—(1) Consumer-to-Shop Clothes Retrieval and (2) InShop Clothes Retrieval—confirm that the proposed loss function not only outperforms the existing loss functions but also achieves the best performance.