Reconstructing the World* in Six Days *(As Captured by the Yahoo 100 Million Image Dataset)

By Jared Heinly, Johannes L. Schönberger, Enrique Dunn, Jan-Michael Frahm. 2015

233 citations recorded by the source corpus.

How researchers see this paper

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

Citations in the corpus

R2D2: Repeatable and Reliable Detector and Descriptor (2019, background): “ints) across images is crucial in many applications such as large-scale visual localization [53, 45], object detection [7], pose estimation [31], Structurefrom-Motion (SfM) [49] and 3D reconstruction [20]. In these applications, extracted keypoints should be sparse, repeatable, and discriminative, i”
Image Matching from Handcrafted to Deep Features: A Survey (2020, methodology): “Next, we use the Visual SfM (Heinly et al. 2015) to recover the camera poses and generate the ground-truth.”
Learning Two-View Correspondences and Geometry Using Order-Aware Network (2019, methodology): “We use the camera poses and sparse models provided by [10] to generate ground-truth.”
Recent developments in large-scale tie-point matching (2016, background): “Modern projects can consist of thousands or even millions of unordered images [16].”
DeepV2D: Video to Depth with Differentiable Structure from Motion (2018, methodology): “This pipeline has shown impressive results on a number of tasks [1, 15] and is often paired with Multi-View Stereo to build a more complete 3D representation [8, 9].”
Comparative Evaluation of Hand-Crafted and Learned Local Features (2017, background): “The descriptor matching uses an efficient GPU implementation, and we use COLMAP [37,40] for the SFM and MVS evaluation, while CMVS [11] is used to cluster the larger datasets into more manageable image clusters for the dense reconstruction.”
Structure-from-Motion Revisited (2016, background): “Oftentimes, the systems fail to register a large fraction of images that empirically should be registrable [20, 30], or the systems produce broken models due to mis-registrations or drift.”
LightGlue: Local Feature Matching at Light Speed (2023, background): “This opens up exciting prospects for deploying deep matchers in latency-sensitive applications like SLAM [45, 5] or re-constructing larger scenes from crowd-sourced data [25, 60, 39, 57].”