Abstract. State-of-the-art automated image orientation (Structure from Motion) and dense image matching (Multiple View Stereo) methods commonly used to produce 3D information from 2D images can generate 3D results – such as point cloud or meshes – of varying geometric and visual quality. Pipelines are generally robust and reliable enough, mostly capable to process even large sets of unordered images, yet the final results often lack completeness and accuracy, especially while dealing with real-world cases where objects are typically characterized by complex geometries and textureless surfaces and obstacles or occluded areas may also occur. In this study we investigate three of the available commonly used open-source solutions, namely COLMAP, OpenMVG+OpenMVS and AliceVision, evaluating their results under diverse large scale scenarios. Comparisons and critical evaluation on the image orientation and dense point cloud generation algorithms is performed with respect to the corresponding ground truth data. The presented FBK-3DOM datasets are available for research purposes.

No abstract

Abstract. Despite the recent success of learning-based monocular depth estimation algorithms and the release of large-scale datasets for training, the methods are limited to depth map prediction and still struggle to yield reliable results in the 3D space without additional scene cues. Indeed, although state-of-the-art approaches produce quality depth maps, they generally fail to recover the 3D structure of the scene robustly. This work explores supervised CNN architectures for monocular depth estimation and evaluates their potential in 3D reconstruction. Since most available datasets for training are not designed toward this goal and are limited to specific indoor scenarios, a new metric, large-scale synthetic benchmark (ArchDepth) is introduced that renders near real-world scenarios of outdoor scenes. A encoder-decoder architecture is used for training, and the generalization of the approach is evaluated via depth inference in unseen views in synthetic and real-world scenarios. The depth map predictions are also projected in the 3D space using a separate module. Results are qualitatively and quantitatively evaluated and compared with state-of-the-art algorithms for single image 3D scene recovery.