Finding local features that are repeatable across multiple views is a cornerstone of sparse 3D reconstruction. The classical image matching paradigm detects keypoints per-image once and for all, which can yield poorly-localized features and propagate large errors to the final geometry. In this paper, we refine two key steps of structure-from-motion by a direct alignment of low-level image information from multiple views: we first adjust the initial keypoint locations prior to any geometric estimation, and subsequently refine points and camera poses as a post-processing. This refinement is robust to large detection noise and appearance changes, as it optimizes a featuremetric error based on dense features predicted by a neural network. This significantly improves the accuracy of camera poses and scene geometry for a wide range of keypoint detectors, challenging viewing conditions, and off-the-shelf deep features. Our system easily scales to large image collections, enabling pixel-perfect crowd-sourced localization at scale. Our code is publicly available at https://github.com/cvg/pixel-perfect-sfm as an add-on to the popular Structure-from-Motion software COLMAP.
Download full-text PDF |
Source |
|---|---|
| http://dx.doi.org/10.1109/TPAMI.2023.3237269 | DOI Listing |
Publication Analysis
Top Keywords
Similar Publications
Single-omics approaches often provide a limited view of complex biological systems, whereas multiomics integration offers a more comprehensive understanding by combining diverse data views. However, integrating heterogeneous data types and interpreting the intricate relationships between biological features-both within and across different data views-remains a bottleneck. To address these challenges, we introduce COSIME (Cooperative Multi-view Integration and Scalable Interpretable Model Explainer).
View Article and Find Full Text PDFMulti-view learning with enhanced multi-weight vector projection support vector machine.
Neural Netw
January 2025
School of Management, Shanghai University of International Business and Economics, Shanghai 201620, China. Electronic address:
Multi-view learning aims on learning from the data represented by multiple distinct feature sets. Various multi-view support vector machine methods have been successfully applied to classification tasks. However, the existed methods often face the problems of long processing time or weak generalization on some complex datasets.
View Article and Find Full Text PDFFace age modulates face ensemble coding.
Vision Res
January 2025
Department of Psychology, College of Education, Hunan Agricultural University.
Research has demonstrated that humans possess the remarkable ability to swiftly extract ensemble statistics, specifically the average identity, from sets of stimuli, such as facial crowds. This phenomenon is known as ensemble perception. Although previous studies have investigated how physiognomic features like gender and race influence face ensemble perception, the impact of face age on face ensemble coding performance remains a relatively unexplored area.
View Article and Find Full Text PDFMultimodal strategies for the implementation of infection prevention and control (IPC) interventions - update of a systematic review for the WHO guidelines on core components of IPC programmes at the facility level.
Clin Microbiol Infect
January 2025
Clinic for Infectious Diseases and Hospital Hygiene, University Hospital Zurich, Zurich, Switzerland; Faculty of Medicine, University of Zurich, Zurich, Switzerland. Electronic address:
Background: Healthcare-associated infections (HAIs) remain a significant challenge worldwide, and the use of multimodal strategies is recommended by the World Health Organization (WHO) to enhance infection prevention.
Objectives: To update the systematic review on facility-level infection prevention and control (IPC) interventions on the WHO Core Component of using multimodal strategies.
Methods: Data Sources: Medline (via PubMed), EMBASE, CINAHL, and the Cochrane library.
Segment Any Leaf 3D: A Zero-Shot 3D Leaf Instance Segmentation Method Based on Multi-View Images.
Sensors (Basel)
January 2025
School of Electronic and Communication Engineering, Sun Yat-sen University, Shenzhen 518000, China.
Exploring the relationships between plant phenotypes and genetic information requires advanced phenotypic analysis techniques for precise characterization. However, the diversity and variability of plant morphology challenge existing methods, which often fail to generalize across species and require extensive annotated data, especially for 3D datasets. This paper proposes a zero-shot 3D leaf instance segmentation method using RGB sensors.
View Article and Find Full Text PDFWant AI Summaries of new PubMed Abstracts delivered to your In-box?
Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!