Instance-level Image Retrieval using Reranking Transformers

Tan, Fuwen; Yuan, Jiangbo; Ordóñez, Vicente

doi:10.1109/iccv48922.2021.01189

Cited by 68 publications

(32 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…3) For a fair comparison, we attach the local branch of the DELG [8] to our global backbone to learn the local DELG features. With these learned local features, we reproduce two re-ranking methods: geometric verification (GV) and Reranking Transformer [45]. Details of the reproduction are provided in the supplementary material.…”

Section: Resultsmentioning

confidence: 99%

“…Although the techniques of global and local representations has progressed significantly, geometric verification remains a de facto solution for image retrieval re-ranking in both conventional [32,33,51] and recent studies [8,30,41,46]. In a recent study, Reranking Transformers (RRT) [45] were proposed as a replacement for geometric verification by leveraging the transformer structure [49]. However, no significant improvement in performance was reported.…”

Section: Related Workmentioning

confidence: 99%

“…Please see the supplementary material for a more detailed explanation of the data selection and sampling process used for the CVNet-Rerank. Finally, we select 1M images from 31k landmarks, and the proposed reranking model is trained for 200 epochs (6.3M steps) for all classes, using a learning rate of 0.00375 and a batch size of 96. r H and p has linearly increase from 0.2 to 1.0 and from 0 to 0. ies [8,13,30,36,45]. We extract global descriptors of three scales:…”

Section: Implementation Detailsmentioning

confidence: 99%

“…Along with deep learning, image retrieval has also advanced significantly. In particular, several studies [8,30,41,45,46,53] have been focused on extracting representative and distinctive features for global and local representations with deep learning. However, geometric verification after local feature matching still plays an essential role in the re-ranking task in image retrieval, despite its drawbacks.…”

Section: Introductionmentioning

confidence: 99%

“…In addition, geometric verification does not include a component that can handle multi-scale operation. Thus, several studies [8,30,32,45] have attempted to solve the scale problem by repeating inference with the image pyramid to extract multi-scale local features. However, this is an extremely expensive process.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Correlation Verification for Image Retrieval

Lee¹,

Seong²,

Lee³

et al. 2022

Preprint

View full text Add to dashboard Cite

Geometric verification is considered a de facto solution for the re-ranking task in image retrieval. In this study, we propose a novel image retrieval re-ranking network named Correlation Verification Networks (CVNet). Our proposed network, comprising deeply stacked 4D convolutional layers, gradually compresses dense feature correlation into image similarity while learning diverse geometric matching patterns from various image pairs. To enable cross-scale matching, it builds feature pyramids and constructs cross-scale feature correlations within a single inference, replacing costly multi-scale inferences. In addition, we use curriculum learning with the hard negative mining and Hide-and-Seek strategy to handle hard samples without losing generality. Our proposed re-ranking network shows state-of-the-art performance on several retrieval benchmarks with a significant margin (+12.6% in mAP on ROxford-Hard+1M set) over state-of-the-art methods. The source code and models are available online: https://github.com/sungonce/CVNet.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Implementation Detailsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Correlation Verification for Image Retrieval

Lee¹,

Seong²,

Lee³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

AttDLNet: Attention-Based Deep Network for 3D LiDAR Place Recognition

Barros

Garrote

Pereira

et al. 2022

Lecture Notes in Networks and Systems

View full text Add to dashboard Cite

Robust and reliable place recognition and loop closure detection in agricultural environments is still an open problem. In particular, orchards are a difficult case study due to structural similarity across the entire field. In this work, we address the place recognition problem in orchards resorting to 3D LiDAR data, which is considered a key modality for robustness. Hence, we propose ORCHNet, a deep-learning-based approach that maps 3D-LiDAR scans to global descriptors. Specifically, this work proposes a new global feature aggregation approach, which fuses multiple aggregation methods into a robust global descriptor. ORCHNet is evaluated on real-world data collected in orchards, comprising data from the summer and autumn seasons. To assess the robustness, We compare ORCHNet with state-ofthe-art aggregation approaches on data from the same season and across seasons. Moreover, we additionally evaluate the proposed approach as part of a localization framework, where ORCHNet is used as a loop closure detector. The empirical results indicate that, on the place recognition task, ORCHNet outperforms the remaining approaches, and is also more robust across seasons. As for the localization, the edge cases where the path goes through the trees are solved when integrating ORCHNet as a loop detector, showing the potential applicability of the proposed approach in this task. The code and dataset will be publicly available at:https://github.com/Cybonic/ORCHNet.git

show abstract

Self-supervised Vision Transformers with Data Augmentation Strategies Using Morphological Operations for Writer Retrieval

Peer

Kleber

Sablatnig

2022

Frontiers in Handwriting Recognition

View full text Add to dashboard Cite

This paper presents an unsupervised approach for writer retrieval based on clustering SIFT descriptors detected at keypoint locations resulting in pseudo-cluster labels. With those cluster labels, a residual network followed by our proposed NetRVLAD, an encoding layer with reduced complexity compared to NetVLAD, is trained on 32×32 patches at keypoint locations. Additionally, we suggest a graph-based reranking algorithm called SGR to exploit similarities of the page embeddings to boost the retrieval performance. Our approach is evaluated on two historical datasets (Historical-WI and HisIR19). We include an evaluation of different backbones and NetRVLAD. It competes with related work on historical datasets without using explicit encodings. We set a new State-of-the-art on both datasets by applying our reranking scheme and show that our approach achieves comparable performance on a modern dataset as well.

show abstract

Instance-level Image Retrieval using Reranking Transformers

Cited by 68 publications

References 41 publications

Correlation Verification for Image Retrieval

Correlation Verification for Image Retrieval

AttDLNet: Attention-Based Deep Network for 3D LiDAR Place Recognition

Self-supervised Vision Transformers with Data Augmentation Strategies Using Morphological Operations for Writer Retrieval

Contact Info

Product

Resources

About