DCR‐Net: Dilated convolutional residual network for fashion image retrieval

Abstract

Fashion image retrieval is an important branch of image retrieval technology.With the rapid development of online shopping, fashion image retrieval technology has made a breakthrough from text-based to content-based. But there is still not a proper deep learning method used for fashion image retrieval. This article proposes a fashion image retrieval framework based on dilated convolutional residual network which consists of two major parts, image feature extraction and feature distance measurement. For image feature extraction, we first extract the shallow features of the input image by a multi-scale convolutional network, and then develop a novel dilated convolutional residual network to obtain the deep features of the image. Finally, the extracted features are transformed into high-dimensional features vector by a binary retrieval vector module. For feature distance measurement, we first use PCA to reduce the dimension of the extracted high-dimensional vectors. Then we propose a mixed distance measurement algorithm combined with cosine distance and Mahalanobis distance to calculate the spatial distance of the feature vectors for similarity ranking, which solves the problems of poor robustness in complex background fashion image retrieval and the inefficiency calculation of Mahalanobis distance. The experimental results show the superiority of our fashion image retrieval framework over existing state-of-the-art methods.

Learn from Restoration: Exploiting Oriented Knowledge Distillation in Self-Supervised Learning for Person Re-Identification

“…Based on these advancements, researchers have concentrated on detailed pedestrian feature learning. This includes areas such as image retrieval [12,13] and pedestrian reconstruction [14]. Further, investigations into fine-grained attribute extraction highlight the trend of in-depth pedestrian feature analysis like body pose estimation [15][16][17], detailed gesture [18] and face sketch analysis [19].…”

SPS-UNet:A Super-pixel Sampling UNet for Extracting Buildings from High-resolution Satellite Images

“…Long et al propose a deep neural network known as fully convolutional network(FCN) [19] that uses convolutional layers instead of fully connected layers to achieve the goal of semantic segmentation; Ronneberger et al propose an other typical network structure known as U-Net based on the FCN [20] that extends the skip connection to compensate for the information loss in the down-sampling process and greatly improves the semantic segmentation accuracy; Badrinarayanan et al propose a SegNet network that records the index information of the max value during max-pooling and then achieves non-linear up-sampling through the corresponding pooling index during decoding; Zhao et al propose a network structure called PSPNet [21] which uses the pyramid pooling module to process the feature information of the backbone in an effort to improve the segmentation accuracy, and this idea has been adopted in many subsequent methods [22][23][24]; Chen et al propose the series of DeepLab networks [25][26][27][28] that use atrous convolutions to enlarge the receptive field. Deep learning has been applied in many fields because of its excellent performance [29][30][31][32][33][34][35].…”

A Thangka cultural element classification model based on self-supervised contrastive learning and MS Triplet Attention

“…In fine-tuning model, we take a two-branch structure. In view of the characteristics of Thangka cultural elements, such as complex picture, complex texture and multi-layer color representation, we use the Local Binary Pattern (LBP) [7][8][9] feature extraction operator to extract the texture features of Thangka cultural elements, so as to enhance the expression ability of fusion features, and thus improve the effect of downstream task. In addition, the attention mechanism of multi-scale feature fusion designed in this paper integrates important features as the feature expression of the final target domain samples, so as to enhance the attention of the effective regions in the proposed features.…”