Retrieval Augmented Classification for Long-Tail Visual Recognition
Open accessThe DOI matches its Crossref record; 28 of the 28 references it cites were found in a registry or an index.
Abstract
We introduce Retrieval Augmented Classification (RAC), a generic approach to augmenting standard image classification pipelines with an explicit retrieval module. RAC consists of a standard base image encoder fused with a parallel retrieval branch that queries a non-parametric external memory of pre-encoded images and associated text snippets. We apply RAC to the problem of long-tail classification and demonstrate a significant improvement over previous state-of-the-art on Places365-LT and iNaturalist-2018 (14.5% and 6.7% respectively), despite using only the training datasets themselves as the external information source. We demonstrate that RAC's retrieval module, without prompting, learns a high level of accuracy on tail classes. This, in turn, frees the base encoder to focus on common classes, and improve its performance thereon. RAC represents an alternative approach to utilizing large, pretrained models without requiring fine-tuning, as well as a first step towards more effectively making use of external memory within common computer vision architectures.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
ImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al. (2020) · arXiv · doi:10.48550/arxiv.2010.11929 · Cited by 47,472
He, H., & Garcia, E. A. (2009) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2008.239 · Cited by 10,359
Representation Learning with Contrastive Predictive Coding
Oord, A. V. D., Li, Y., & Vinyals, O. (2018) · arXiv · doi:10.48550/arxiv.1807.03748 · Cited by 6,242
Borderline-SMOTE : A New Over-Sampling Method in Imbalanced Data Sets Learning
Han, H., Wang, W.-Y., & Mao, B.-H. (2005) · Lecture notes in computer science · doi:10.1007/11538059_91 · Cited by 4,331 · cited 2×
“Resampling and Logit Adjustment Over-sampling sparse classes [3, 19] is one of the oldest approaches to addressing distribution bias, but one that is still in common use.”
BackgroundBillion-scale similarity search with GPUs
Johnson, J., Douze, M., & Jegou, H. (2021) · IEEE Transactions on Big Data · doi:10.1109/tbdata.2019.2921572 · Cited by 4,228
Billion-scale similarity search with GPUs
Johnson, J., Douze, M., & Jégou, H. (2017) · arXiv · doi:10.48550/arxiv.1702.08734 · Cited by 4,140 · cited 1×
“For fast querying of the index, we make use of the FAISS implementation [23] of the Hierarchical Navigable Small World (HNSW) approximate k-NN lookup [35].”
MethodPlaces: A 10 Million Image Database for Scene Recognition
Zhou, B., Lapedriza, A., Khosla, A., Oliva, A., & Torralba, A. (2018) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2017.2723009 · Cited by 4,042
Transformers in Vision: A Survey
Khan, S., Naseer, M., Hayat, M., Zamir, S. W., Khan, F. S., & Shah, M. (2022) · ACM Computing Surveys · doi:10.1145/3505244 · Cited by 3,595 · cited 1×
“Large Transformer [48] models have arrived in Computer Vision, with parameter counts and pretraining dataset size increasing rapidly [11,26,34,42,44,53].”
MethodTransformers in Vision: A Survey
Khan, S., Naseer, M., Hayat, M., Zamir, S. W., Khan, F. S., & Shah, M. (2021) · arXiv · doi:10.48550/arxiv.2101.01169 · Cited by 3,594
Dense passage retrieval for open-domain question answering
Karpukhin, V., Oguz, B., Min, S., Lewis, P., Wu, L., Edunov, S., Chen, D., & Yih, W.-T. (2020) · Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2020.emnlp-main.550 · Cited by 3,575
On calibration of modern neural networks
Guo, C., Pleiss, G., Sun, Y., & Weinberger, K. Q. (2017) · arXiv · doi:10.48550/arxiv.1706.04599 · Cited by 3,435
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W.-T., Rocktäschel, T., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.11401 · Cited by 3,101 · cited 1×
“Recently, in the NLP domain, several works have proposed the augmentation of large language models with a non-parametric memory to allow explicit access to external data [18, 30].”
BackgroundClass-Balanced Loss Based on Effective Number of Samples
Cui, Y., Jia, M., Lin, T.-Y., Song, Y., & Belongie, S. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00949 · Cited by 3,094 · cited 1×
“Instead, a small number of samples may still define a sufficiently clear decision boundary if the volume of semantic space covered by that class is small and distinct [7], and hence in a truly balanced model, both inter- and intra-class distributions must be considered.”
BackgroundRevisiting Unreasonable Effectiveness of Data in Deep Learning Era
Sun, C., Shrivastava, A., Singh, S., & Gupta, A. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.97 · Cited by 2,484
Malkov, Y. A., & Yashunin, D. A. (2020) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2018.2889473 · Cited by 1,904
Dataset Shift in Machine Learning
(2008) · doi:10.7551/mitpress/9780262170055.001.0001 · Cited by 1,553 · cited 1×
“Long-tail learning is a well-studied [2, 20, 39] instance of the more general label shift problem [41], where the shift is static and known during both training and testing.”
BackgroundThe iNaturalist Species Classification and Detection Dataset
Van Horn, G., Mac Aodha, O., Song, Y., Cui, Y., Sun, C., Shepard, A., Adam, H., Perona, P., & Belongie, S. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00914 · Cited by 1,451
Graves, A., Wayne, G., & Danihelka, I. (2014) · arXiv · doi:10.48550/arxiv.1410.5401 · Cited by 1,058 · cited 1×
“External Memory One of the first models to successfully combine deep networks with external memory was the Neural Turing Machine [15].”
MethodBig Transfer (BiT): General Visual Representation Learning
Kolesnikov, A., Beyer, L., Zhai, X., Puigcerver, J., Yung, J., Gelly, S., & Houlsby, N. (2019) · arXiv · doi:10.48550/arxiv.1912.11370 · Cited by 1,039
Big Transfer (BiT): General Visual Representation Learning
Kolesnikov, A., Beyer, L., Zhai, X., Puigcerver, J., Yung, J., Gelly, S., & Houlsby, N. (2020) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-58558-7_29 · Cited by 1,039 · cited 1×
“We make use of IM21k to expand the index in some experiments, and use the variant introduced in [28].”
MethodTransformer Interpretability Beyond Attention Visualization
Chefer, H., Gur, S., & Wolf, L. (2021) · 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr46437.2021.00084 · Cited by 903
Zhai, X., Kolesnikov, A., Houlsby, N., & Beyer, L. (2022) · 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52688.2022.01179 · Cited by 891
When does label smoothing help?
Müller, R., Kornblith, S., & Hinton, G. (2019) · arXiv · doi:10.48550/arxiv.1906.02629 · Cited by 859 · cited 1×
“In our experiments, this calibration is achieved through label smoothing [37], which has been shown to implicitly calibrate neural networks [45].”
MethodLatent retrieval for weakly supervised open domain question answering
Lee, K., Chang, M.-W., & Toutanova, K. (2019) · Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/p19-1612 · Cited by 771 · cited 1×
“Latent retrieval has been applied to textual open-domain QA [25, 29].”
MethodHow to train your ViT? Data, Augmentation, and Regularization in Vision Transformers
Steiner, A., Kolesnikov, A., Zhai, X., Wightman, R., Uszkoreit, J., & Beyer, L. (2021) · arXiv · doi:10.48550/arxiv.2106.10270 · Cited by 292 · cited 1×
“Unless otherwise specified, E(·) is a ViT-B-16, pretrained on ImageNet following [43].”
BackgroundGeneralization through Memorization: Nearest Neighbor Language Models
Khandelwal, U., Levy, O., Jurafsky, D., Zettlemoyer, L., & Lewis, M. (2019) · arXiv · doi:10.48550/arxiv.1911.00172 · Cited by 285
Editing Factual Knowledge in Language Models
De Cao, N., Aziz, W., & Titov, I. (2021) · Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/2021.emnlp-main.522 · Cited by 180 · cited 1×
“The distributed representations learned by such models result in significant performance gains on a range of tasks, however come with the drawback of storing world knowledge implicitly within their parameters, making post-hoc modification [8] and interpretability [4] challenging.”
Background
Cited by
Retrieval-Augmented Egocentric Video Captioning
Xu, J., Huang, Y., Hou, J., Chen, G., Zhang, Y., Feng, R., & Xie, W. (2024) · arXiv · doi:10.48550/arxiv.2401.00789 · Cited by 24
Retrieval-Augmented Few-Shot Medical Image Segmentation With Foundation Models
Zhao, L., Chen, X., Chen, E. Z., Liu, Y., Chen, T., & Sun, S. (2025) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2025.3568479 · Cited by 19
“In addition, retrieval-augmented customization (REACT) leverages relevant image–text pairs from a vast web-scale database to customize visual models for specific target domains, achieving significant performance across tasks while minimizing the need for extensive retraining [36].”
Memories are One-to-Many Mapping Alleviators in Talking Face Generation
Tang, A., He, T., Tan, X., Ling, J., & Song, L. (2022) · arXiv · doi:10.48550/arxiv.2212.05005 · Cited by 12
Survey on Memory-Augmented Neural Networks: Cognitive Insights to AI Applications
Khosla, S., Zhu, Z., & He, Y. (2023) · arXiv · doi:10.48550/arxiv.2312.06141 · Cited by 3
“Retrieval Augmented Classification (RAC) [7] tackles the problem of long-tail classification by augmenting a standard image encoder with a retrieval module.”
Background
Metadata
- Authors
- Long, A., Yin, W., Ajanthan, T., Nguyen, V., Purkait, P., Garg, R., Blair, A., Shen, C., & van den Hengel, A.
- Venue
- 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
- Publisher
- IEEE
- Published
- 2022-06
- Pages
- 6949-6959
- ISSN
- 1063-6919 (ISSN-L)
- OpenAlex
- W4312920106
- Cited by
- 80
- References
- 28
- FWCI
- 6.18
- Type
- conference paper
- Language
- en
- Open access
- green
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.