Image representation methods based on deep convolutional neural networks (CNNs) have achieved state-of-the-art performance in various computer vision tasks, such as image retrieval and person re-identification. We recognize that more discriminative feature embeddings can be learned with supervised deep metric learning and hand crafted features for image retrieval and similar applications. In this paper, we propose a new supervised deep feature embedding with a hand crafted feature model. To fuse hand crafted feature information into CNNs and realize feature embeddings, a general fusion unit is proposed (called Fusion-Net). We also define a network loss function with image label information to realize supervised deep metric learning. Our extensive experimental results on the Stanford Online Products’ data set and the In-shop Clothes Retrieval data set demonstrate that our proposed methods outperform existing state-of-the-art methods of image retrieval with a large margin. Moreover, we also explore the applications of the proposed methods in person re-identification and vehicle re-identification; both experimental results demonstrate the effectiveness and efficiency of the proposed methods.

Recent methods for deep metric learning have been focusing on designing different contrastive loss functions between positive and negative pairs of samples so that the learned feature embedding is able to pull positive samples of the same class closer and push negative samples from different classes away from each other. In this work, we recognize that there is a significant semantic gap between features at the intermediate feature layer and class labels at the final output layer. To bridge this gap, we develop a contrastive Bayesian analysis to characterize and model the posterior probabilities of image labels conditioned by their features similarity in a contrastive learning setting. This contrastive Bayesian analysis leads to a new loss function for deep metric learning. To improve the generalization capability of the proposed method onto new classes, we further extend the contrastive Bayesian loss with a metric variance constraint. Our experimental results and ablation studies demonstrate that the proposed contrastive Bayesian metric learning method significantly improves the performance of deep metric learning in both supervised and pseudo-supervised scenarios, outperforming existing methods by a large margin.

The block-based representation learning method has been proven to be a very effective method for person re-identification (Re-ID), but the features extracted by the existing block-based approach tend to have a high correlation among different blocks. Also, these methods perform less well for persons with large posture changes. Thus, Part-based Nondirect Coupling (PNC) representation learning method is proposed by introducing a similarity measure loss to constrain features of different blocks. Moreover, Part-based Nondirect Coupling Embedded GAN (PGAN) method is proposed, which aims to extract more common features of different postures of a same person. In this way, the extracted features of the network are robust for posture changes of a person, and there are no auxiliary pose information and additional computational cost required in the test stage. Experimental results on public datasets show that our proposed method achieves good performances, especially, it outperforms the state-of-the-art GAN-based methods for person Re-ID.

Deep feature embedding aims to learn discriminative features or feature embeddings for image samples which can minimize their intra-class distance while maximizing their inter-class distance. Recent state-of-the-art methods have been focusing on learning deep neural networks with carefully designed loss functions. In this work, we propose to explore a new approach to deep feature embedding. We learn a graph neural network to characterize and predict the local correlation structure of images in the feature space. Based on this correlation structure, neighboring images collaborate with each other to generate and refine their embedded features based on local linear combination. Graph edges learn a correlation prediction network to predict the correlation scores between neighboring images. Graph nodes learn a feature embedding network to generate the embedded feature for a given image based on a weighted summation of neighboring image features with the correlation scores as weights. Our extensive experimental results under the image retrieval settings demonstrate that our proposed method outperforms the state-of-the-art methods by a large margin, especially for top-1 recalls.