HCPM: Hierarchical Candidates Pruning for Efficient Detector-Free Matching
Not on the map. Why?
Abstract (arXiv)
Deep learning-based image matching methods play a crucial role in computer vision, yet they often suffer from substantial computational demands. To tackle this challenge, we present HCPM, an efficient and detector-free local feature-matching method that employs hierarchical pruning to optimize the matching pipeline. In contrast to recent detector-free methods that depend on an exhaustive set of coarse-level candidates for matching, HCPM selectively concentrates on a concise subset of informative candidates, resulting in fewer computational candidates and enhanced matching efficiency. The method comprises a self-pruning stage for selecting reliable candidates and an interactive-pruning stage that identifies correlated patches at the coarse level. Our results reveal that HCPM significantly surpasses existing methods in terms of speed while maintaining high accuracy. The source code will be made available upon publication.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 35 of its 40 references that are in the map. First posted on arXiv 19 Mar 2024.
Builds on
35 of 40 references.
References · 35
- 2017Attention Is All You Needcited by 59,932
- 2015Deep Residual Learning for Image Recognitioncited by 51,809
- 2016Feature Pyramid Networks for Object Detectioncited by 5,021
- 2022FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awarenesscited by 3,611
- 2020Longformer: The Long-Document Transformercited by 3,362
- 2016Categorical Reparameterization with Gumbel-Softmaxcited by 2,708
- 2020Transformers are RNNs: Fast Autoregressive Transformers with Linear Attentioncited by 1,979
- 2020Big Bird: Transformers for Longer Sequencescited by 1,673
- 2020Reformer: The Efficient Transformercited by 1,658
- 2020Rethinking Attention with Performerscited by 1,584
- 2020Linformer: Self-Attention with Linear Complexitycited by 1,462
- 2019SuperGlue: Learning Feature Matching with Graph Neural Networkscited by 1,437
23 more
- 2017SuperPoint: Self-Supervised Interest Point Detection and Descriptioncited by 1,313
- 2016The Concrete Distribution: A Continuous Relaxation of Discrete Random Variablescited by 1,135
- 2021LoFTR: Detector-Free Local Feature Matching with Transformerscited by 974
- 2018MegaDepth: Learning Single-View Depth Prediction from Internet Photoscited by 731
- 2023LightGlue: Local Feature Matching at Light Speedcited by 583
- 2018Efficient Attention: Attention with Linear Complexitiescited by 366
- 2017HPatches: A benchmark and evaluation of handcrafted and learned local descriptorscited by 261
- 2019R2D2: Repeatable and Reliable Detector and Descriptorcited by 206
- 2021Adaptive Token Sampling For Efficient Vision Transformerscited by 181
- 2022ASpanFormer: Detector-Free Image Matching with Adaptive Span Transformercited by 170
- 2021Patch Slimming for Efficient Vision Transformerscited by 131
- 2022QuadTree Attention for Vision Transformerscited by 129
- 2020Patch2Pix: Epipolar-Guided Pixel-Level Correspondencescited by 128
- 2022MatchFormer: Interleaving Attention in Transformers for Feature Matchingcited by 126
- 2021Learning to Match Features with Seeded Graph Matching Networkcited by 92
- 2023Joint Token Pruning and Squeezing Towards More Aggressive Compression of Vision Transformerscited by 79
- 2022ClusterGNN: Cluster-based Coarse-to-Fine Graph Neural Network for Efficient Feature Matchingcited by 62
- 2023Dynamic Token Pruning in Plain Vision Transformers for Semantic Segmentationcited by 61
- 2023SparseViT: Revisiting Activation Sparsity for Efficient High-Resolution Vision Transformercited by 60
- 2016Quad-networks: unsupervised learning to rank for interest point detectioncited by 36
- 2022Unleashing Vanilla Vision Transformer with Masked Image Modeling for Object Detectioncited by 33
- 2022Adaptive Assignment for Geometry Aware Local Feature Matchingcited by 30
- 2022Guide Local Feature Matching by Overlap Estimationcited by 24
Built on by
Citing papers per year: 2024: 1.
1 paper.