Online Submission ID: 0184 Photo Tourism: Exploring Photo Collections in 3D

By Noah Snavely, Steven M. Seitz, Richard Szeliski. 2006

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

YFCC100M (2015, background): “In the past, geotagged media has been used to address a variety of research endeavors, such as location estimation [9], event detection [15], ﬁnding canonical views of places [4], and visual reconstruction of the world [17].”
Depth Map Prediction from a Single Image using a Multi-Scale Deep Network (2014, background): “[20] match across views of many uncalibrated consumer photographs of the same scene to create accurate 3D reconstructions of common landmarks.”
IM2GPS: estimating geographic information from a single image (2008, methodology): “Similar feature-based geometric matching approaches have also been successfully applied to co-registering online photographs of famous landmarks for browsing [14] and summarization [13],”
Learning to compare image patches via convolutional neural networks (2015, methodology): “This allows to further improve the performance of our method simply by enriching this database with more samples (as software for automatically generating such samples is readily available [21]).”
First-person hyper-lapse videos (2014, methodology): “The original algorithm [Snavely et al. 2006] computes tracks by connecting all pairwise matching feature points with common endpoints.”
Scaffold-GS: Structured 3D Gaussians for View-Adaptive Rendering (2023, methodology): “Ini-tialized from point clouds derived from Structure from Motion (SfM) [42], this method optimizes a set of 3D Gaus-sians to represent the scene.”
Robust monocular SLAM in dynamic environments (2013, methodology): “The structure from motion (SfM) technique provides a fundamental framework for offline 3D structure reconstruction from photo collections [37] or video sequences [43].”
Clues from the beaten path: Location estimation with bursty sequences of tourist photos (2011, methodology): “When available, georeference data from GPS sensors can help automate some aspects of this organization; however, GPS falls short for images taken indoors, and a mere 3% of the billions of existing consumer photos online actually have a GPS record [1].”
Using Unmanned Aerial Vehicles (UAV) for High-Resolution Reconstruction of Topography: The Structure from Motion Approach on Coastal Environments (2013, background): “[10], Snavely [11], and Ullman [12] for a detailed discussion on SfM.”
Robust Global Translations with 1DSfM (2014, methodology): “Because ground truth positions are usually unavailable for large-scale SfM problems, we show our method gives similar results to a sequential SfM system based on Bundler [19], but in much less time.”
Global Motion Estimation from Point Matches (2012, methodology): “We compare our running time to VisualSFM [24], which is a highly optimized GPU implementation of the Photo Tourism [27]’s sequential algorithm.”
Continuous 3D Perception Model with Persistent State (2025, unknown): “Notable examples include traditional methods such as Structure from Motion (SfM) [1, 2, 28, 47, 65, 66, 70, 71, 80] and Simultaneous Localization and Mapping (SLAM) [11, 16, 19, 22, 37, 51, 53], as well as more recent approaches such as Neural Radiance Fields (NeRF) [12, 24, 49, 50, 84] and 3D…”
Learning Local Image Descriptors (2007, methodology): “Interest point detectors and descriptors have become popular for obtaining image to image correspondence for 3D reconstruction [17, 14], searching databases of photographs [10] and as a ﬁrst stage in object or place recognition [8, 13].”
The New Data and New Challenges in Multimedia Research (2015, background): “In the past, geotagged media has been used to address a variety of research endeavors, such as location estimation [19], event detection [20], finding canonical views of places [21], and visual reconstruction of the world [22].”
MegaDepth: Learning Single-View Depth Prediction from Internet Photos (2018, background): “Estimating geometry from Internet photo collections has been an active research area for a decade, with advances in both structure from motion [35, 2, 37, 30] and multi-view stereo [14, 9, 32].”
D-NeRF: Neural Radiance Fields for Dynamic Scenes (2020, background): “…ﬁeld of neural rendering, which learn scene representations encoding both geometry and appearance [26, 23, 19, 50, 29, 35], have achieved re-sults that largely surpass those of traditional Structure-from-Motion [14, 41, 38], light-ﬁeld photography [18] and image-based rendering approaches [5].”
Total Recall: Automatic Query Expansion with a Generative Feature Model for Object Retrieval (2007, methodology): “With these more sophisticated models we could imagine returning a synthesis of the queried object directly rather than a set of matching images [19].”
Straightforward reconstruction of 3D surfaces and topography with a camera: Accuracy and geoscience application (2012, methodology): “…(http://blog.neonascent.net/archives/ bundler-photogrammetry-package/ by J. Harle) that combines the commonly used SfM application ‘ Bundler ’ [ Snavely et al. , 2006, 2008a, 2008b], with the MVS dense matcher ‘ PMVS2 ’ (Patch-based MultiView Stereo) [ Furukawa and Ponce , 2007, 2010].”