… The sphere’s radius is equal to the evaluation tolerance t. Reconstructed points outside all …
To measure a voxel’s completeness, we use the ground truth points in it and all reconstructed …

… (fcns), which output the class likelihoods for an entire image at once. fcns have become a …
Still, we find that when applied to remote sensing data with many small objects, fcns tend to …

OBJECTIVE Poor outcomes have been reported in patients with type 2 diabetes and
coronavirus disease 2019 (COVID-19); thus, it is mandatory to explore novel therapeutic …

… In big datasets where the angle criteria still produces too many views, we propose to randomly
pick a subset S of views within this selection only if the runtime performance is preferred …

… Apart form our standard setting of N = 5 views for DTU’s evaluation set [1], we also evaluate
the performance when N = 2, 3, 6. Using more views is known to improve performance, eg. …

This paper describes a deep learning approach to semantic segmentation of very high
resolution (aerial) images. Deep neural architectures hold the promise of end-to-end learning …

… s, we downsample the original S2 data, by first blurring it with a Gaussian filter of standard
deviation σ = 1 / s … downsample by averaging over s × s windows, with s = 2 respectively s = 6 . …

Mixed reality headsets, such as the Microsoft HoloLens 2, are powerful sensing devices
with integrated compute capabilities, which makes it an ideal platform for computer vision …

We present IterMVS, a new data-driven method for high-resolution multi-view stereo. We
propose a novel GRU-based estimator that encodes pixel-wise probability distributions of depth …

… Therefore, we opt for the ScanNet [11] dataset’s official training split to train and validate
our models. For testing, we use ScanNet’s 100 sequence official test split and a diverse …