Cross-Batch Memory for Embedding Learning

Wang, Xun; Zhang, Haozhi; Huang, Weilin; Scott, Matthew R.

doi:10.1109/cvpr42600.2020.00642

Cited by 206 publications

(164 citation statements)

References 38 publications

Supporting

Mentioning

162

Contrasting

Order By: Relevance

“…Moreover, in those experiments, we use the simpler data augmentation for faster convergence since our focus is analysing the key components, instead of comparing with existing methods. Specifically, we crop a random size of 224×224 [68] is marked with '*' because it exploits information across mini-batch tasks. The '-' denotes the corresponding results are not reported in the original paper.…”

Section: Training and Optimisation Settingsmentioning

confidence: 99%

“…Additionally, SoftMax norm and SoftTriple [39] are theoretically non-scalable to extremely large dataset because they use multiple proxies to represent one class. XBM [68] exploits extra information across mini-batch tasks. Some other methods, e.g., Margin [71], Divide & Conquer [43], FastAP [4] and MIC [41], use ResNet-50 [13] as the backbone network.…”

Section: Comparison With Recent Baselinesmentioning

confidence: 99%

See 1 more Smart Citation

Ranked List Loss for Deep Metric Learning

Wang

Yang

Kodirov³

et al. 2021

IEEE Trans. Pattern Anal. Mach. Intell.

View full text Add to dashboard Cite

The objective of deep metric learning (DML) is to learn embeddings that can capture semantic similarity and dissimilarity information among data points. Existing pairwise or tripletwise loss functions used in DML are known to suffer from slow convergence due to a large proportion of trivial pairs or triplets as the model improves. To improve this, ranking-motivated structured losses are proposed recently to incorporate multiple examples and exploit the structured information among them. They converge faster and achieve state-of-the-art performance. In this work, we unveil two limitations of existing ranking-motivated structured losses and propose a novel ranked list loss to solve both of them. First, given a query, only a fraction of data points is incorporated to build the similarity structure. Consequently, some useful examples are ignored and the structure is less informative. To address this, we propose to build a set-based similarity structure by exploiting all instances in the gallery. The learning setting can be interpreted as few-shot retrieval: given a mini-batch, every example is iteratively used as a query, and the rest ones compose the galley to search, i.e., the support set in few-shot setting. The rest examples are split into a positive set and a negative set. For every mini-batch, the learning objective of ranked list loss is to make the query closer to the positive set than to the negative set by a margin. Second, previous methods aim to pull positive pairs as close as possible in the embedding space. As a result, the intraclass data distribution tends to be extremely compressed. In contrast, we propose to learn a hypersphere for each class in order to preserve useful similarity structure inside it, which functions as regularisation. Extensive experiments demonstrate the superiority of our proposal by comparing with the state-of-the-art methods on the fine-grained image retrieval task. Our source code is available online: https://github.com/XinshaoAmos Wang/Ranked-List-Loss-for-DML.

show abstract

Section: Training and Optimisation Settingsmentioning

confidence: 99%

Section: Comparison With Recent Baselinesmentioning

confidence: 99%

Ranked List Loss for Deep Metric Learning

Wang

Yang

Kodirov³

et al. 2021

IEEE Trans. Pattern Anal. Mach. Intell.

View full text Add to dashboard Cite

show abstract

“…Therefore, by minimizing this loss, we force the network to assign high similarity values on the positive pairs, i.e., the images originating from the same cell, and low similarity values on the negative pairs, i.e., the images that come from different cells. To utilize more negative samples during the loss calculation, we employ a crossbatch memory bank [45,48]. Additionally, to eliminate the bias from the cells that contain many images, in each training epoch, we sample one image pair from each cell.…”

Section: Training Processmentioning

confidence: 99%

Leveraging EfficientNet and Contrastive Learning for Accurate Global-scale Location Estimation

Kordopatis-Zilos

Galopoulos

Papadopoulos

et al. 2021

Proceedings of the 2021 International Conference on Multimedia Retrieval

View full text Add to dashboard Cite

In this paper, we address the problem of global-scale image geolocation, proposing a mixed classification-retrieval scheme. Unlike other methods that strictly tackle the problem as a classification or retrieval task, we combine the two practices in a unified solution leveraging the advantages of each approach with two different modules. The first leverages the EfficientNet architecture to assign images to a specific geographic cell in a robust way. The second introduces a new residual architecture that is trained with contrastive learning to map input images to an embedding space that minimizes the pairwise geodesic distance of same-location images. For the final location estimation, the two modules are combined with a search-within-cell scheme, where the locations of most similar images from the predicted geographic cell are aggregated based on a spatial clustering scheme. Our approach demonstrates very competitive performance on four public datasets, achieving new state-of-the-art performance in fine granularity scales, i.e., 15.0% at 1km range on Im2GPS3k. CCS CONCEPTS• Computing methodologies → Computer vision problems;• Information systems → Geographic information systems.

show abstract

“…Recent work in deep metric learning has introduced a number of training objectives with state of the art performance on computer vision tasks (Kim et al, 2020;Wang et al, 2019). Unfortunately, many of these objectives scale linearly with the number K of classes considered due to a costly linear projection onto R K .…”

Section: Scalable Deep Metric Learning Lossesmentioning

confidence: 99%

A Deep Metric Learning Approach to Account Linking

Khan¹,

Fleming²,

Schofield³

et al. 2021

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Langua

View full text Add to dashboard Cite

We consider the task of linking social media accounts that belong to the same author in an automated fashion on the basis of the content and metadata of their corresponding document streams. We focus on learning an embedding that maps variable-sized samples of user activity-ranging from single posts to entire months of activity-to a vector space, where samples by the same author map to nearby points. The approach does not require humanannotated data for training purposes, which allows us to leverage large amounts of social media content. The proposed model outperforms several competitive baselines under a novel evaluation framework modeled after established recognition benchmarks in other domains. Our method achieves high linking accuracy, even with small samples from accounts not seen at training time, a prerequisite for practical applications of the proposed linking framework.

show abstract

Cross-Batch Memory for Embedding Learning

Cited by 206 publications

References 38 publications

Ranked List Loss for Deep Metric Learning

Ranked List Loss for Deep Metric Learning

Leveraging EfficientNet and Contrastive Learning for Accurate Global-scale Location Estimation

A Deep Metric Learning Approach to Account Linking

Contact Info

Product

Resources

About