Divide&Classify: Fine-Grained Classification for City-Wide Visual Place Recognition
Gabriele Trivigno, Gabriele Moreno Berton, Juan Aragon, Barbara Caputo, Carlo Masone
摘要
Visual Place recognition is commonly addressed as an image retrieval problem. However, retrieval methods are impractical to scale to large datasets, densely sampled from city-wide maps, since their dimension impact negatively on the inference time. Using approximate nearest neighbour search for retrieval helps to mitigate this issue, at the cost of a performance drop. In this paper we investigate whether we can effectively approach this task as a classification problem, thus bypassing the need for a similarity search. We find that existing classification methods for coarse, planet-wide localization are not suitable for the fine-grained and city-wide setting. This is largely due to how the dataset is split into classes, because these methods are designed to handle a sparse distribution of photos and as such do not consider the visual aliasing problem across neighbouring classes that naturally arises in dense scenarios. Thus, we propose a partitioning scheme that enables a fast and accurate inference, preserving a simple learning procedure, and a novel inference pipeline based on an ensemble of novel classifiers that uses the prototypes learned via an angular margin loss. Our method, Divide&Classify (D&C), enjoys the fast inference of classification solutions and an accuracy competitive with retrieval methods on the fine-grained, city-wide setting. Moreover, we show that D&C can be paired with existing retrieval pipelines to speed up computations by over 20 times while increasing their recall, leading to new state-of-the-art results. Code is available at https://github.com/ga1i13o/Divide-and-Classify
问 Lune
几分钟读透这篇 Paper,不用再花一下午。
Lune 能读完全文,也能读引用它的顶会 Paper,讲清它的结论、证据和局限,每一点都注明出自哪篇 Paper。不用硬啃,也不靠猜。
脉络
- Rethinking Visual Geo-localization for Large-Scale Applications 引用了 Scalable Place Recognition Under Appearance Change for Autonomous Driving
- Rethinking Visual Geo-localization for Large-Scale Applications 引用了 Stochastic Attraction-Repulsion Embedding for Large Scale Image Localization
它引用的 Paper9
- Understanding Contrastive Representation Learning through Alignment and Uniformity on the HypersphereTongzhou Wang, Phillip IsolaICML 2020 · 2,360 次引用
- Rethinking Visual Geo-localization for Large-Scale ApplicationsGabriele Moreno Berton, Carlo Masone, Barbara CaputoCVPR 2022 · 235 次引用
- TransVPR: Transformer-Based Place Recognition with Multi-Level Attention AggregationRuotong Wang, Yanqing Shen, Weiliang Zuo, Sanping Zhou 等CVPR 2022 · 167 次引用
- Stochastic Attraction-Repulsion Embedding for Large Scale Image LocalizationLiu Liu, Hongdong Li, Yuchao DaiICCV 2019 · 123 次引用
- Scalable Place Recognition Under Appearance Change for Autonomous DrivingDzung A. Doan, Yasir Latif, Tat-Jun Chin, Yu Liu 等ICCV 2019 · 81 次引用
相关 Paper
- EigenPlaces: Training Viewpoint Robust Models for Visual Place RecognitionGabriele Moreno Berton, Gabriele Trivigno, Barbara Caputo, Carlo MasoneICCV 2023 · 141 次引用
- MutualVPR: A Mutual Learning Framework for Resolving Supervision Inconsistencies via Adaptive ClusteringQiwen Gu, Xufei Wang, Junqiao Zhao, Siyue Tao 等NeurIPS 2025 · 6 次引用
- Data-Efficient Large Scale Place Recognition with Graded Similarity SupervisionMaria Leyva-Vallina, Nicola Strisciuglio, Nicolai PetkovCVPR 2023
- Efficient Large-scale Localization by Global Instance RecognitionFei Xue, Ignas Budvytis, Daniel Olmeda Reino, Roberto CipollaCVPR 2022 · 19 次引用
- SAGE: Spatial-visual Adaptive Graph Exploration for Efficient Visual Place RecognitionShunpeng Chen, Changwei Wang, Rongtao Xu, Xingtian Pei 等ICLR 2026 · 6 次引用
