Exploiting Best Practice of Deep CNNs Features for National Costume Image Retrieval

Abstract

Convolutional neural networks (CNNs) have recently achieved remarkable success with superior performances in computer vision applications. In most CNN-based image retrieval methods, deep CNNs features are verified as discriminative descriptors for effective image representation. This paper exploits the best practice for CNNs application to national costume image retrieval. Several important aspects that affect the discriminative ability of deep CNNs features are investigated thoroughly, including layers selection, aggregation and weighting methods. Firstly, an effective weighting method for sum-pooling features aggregation is given, which is more suitable for national costume image than some typical aggregation methods such as SPoC and SCDA. Secondly, in view of the complementary strengths, compact multi-layer CNN features combined with low dimensions are proposed and proven to be effective for national costume expression. Finally, a re-ranking strategy of diffusion process is applied to further enhance the performance for national costume images retrieval. The experimental results show that the proposed method outperforms the existing methods remarkably, which will provide some new research ideas and technical references for researchers in the field of national costume image retrieval.

Convolutional neural network application for supply–demand matching in Zhuang ethnic clothing image classification

“…This method effectively enhanced the model’s capability to classify edge cases and complex samples by actively generating challenging samples, strengthening overall robustness. Juxiang Zhou & Gan (2018) 18 , focusing on the characteristics of ethnic clothing images, optimized CNN using a feature aggregation weighting method. They further improved the performance of ethnic clothing image retrieval by employing a reordering strategy based on diffusion processes, providing new insights and technical references for researchers in the field of ethnic clothing image retrieval.…”

Research on Digital Construction and Design of Minority Clothing Based on Multivariate Statistical Analysis

“…Literature [12] combined with computer software modeling technology through human-computer interaction in the 3D game scene model on the optimization of the character clothing design, the network character modeling and design process in-depth study, and the application of ethnic clothing and game clothing design, to promote the development of clothing and online games. Literature [13] combines hierarchical analysis as well as clustering weighting methods to assign weights to national costume images, combines convolutional neural networks to mine costumes, improves the performance of national costume image retrieval using application diffusion methods, and utilizes best practices for national costume image retrieval. Literature [14] utilizes the data mining and the things technology of the Internet to construct an intelligent clothing design system with intelligent human-computer interaction concepts and utilizes constrained spectral clustering algorithm to improve the data processing effect of the ethnic minority clothing design, which can effectively improve the design effect of the ethnic minority clothing.…”