A Benchmark and a Baseline for Robust Multi-view Depth Estimation

Philipp, Schröppel,; Bechtold, Jan; Amiranashvili, Artemij; Brox, Thomas

doi:10.1109/3dv57658.2022.00074

Cited by 4 publications

(4 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For MVD, we compare with famous classic COLMAP [55], [56] and Deep learning based methods, such as MVS-Net [33], Vis-MVSSNet [37], MVS2D [57], DeMon [28], DeepV2D [30], Robust MVD baseline [31], DUSt3R [42] and more. In particular, the latest fully supervised model DUSt3R [42] supports the closest to us.…”

Section: Methodsmentioning

confidence: 99%

“…Not only operating on pair of images, DeepTAM [29] and DeepV2D [30] process more images with alternating mapping and tracking modules. However, according to [31], such methods are overfitted to the trained scale and camera parameters. Which are difficult to generalize to arbitrary real-world applications.…”

Section: Multi-view Depth Estimationmentioning

confidence: 99%

“…In the MVD test, we follow Ro-bustMVD [31] to use relative depth error and inlier ratio to metric the estimation quality. Since our model does not input both ground truth (GT) depth and inlier, for fair evaluation we follow DUSt3R [42] to align the depth to GT scale with median value.…”

Section: ) Evaluation Metricsmentioning

confidence: 99%

“…One of the most important components of SceneFactory is the depth estimator. We follow the Multiview Depth Estimation (MVD) benchmark, RobustMVD [31], to evaluate the depth result on 5 widely used datasets (KITTI, ScanNet, ETH3D, DTU, and Tanks&Temple). Because our model also support intrinsic-free, which is not listed in [31], so besides of the whole benchmark, we find DUSt3R [42] as the most similar work to follow for comparison.…”

Section: B Effect Of Multiview Detph Estimationmentioning

confidence: 99%

See 3 more Smart Citations

A Unified Inference Framework for Multiple Imputation Using Martingales

Guan¹,

Yang²

2024

STAT SINICA

View full text Add to dashboard Cite

Multiple imputation is widely used to handle missing data. Although Rubin's combining rule is simple, it is not clear whether or not the standard multiple imputation inference is consistent when coupled with the commonly-used full sample estimators. This article establishes a unified martingale representation of multiple imputation for a wide class of asymptotically linear full sample estimators. This representation invokes the wild bootstrap inference to provide consistent variance estimation under the correct specification of the imputation models. As a motivating application, we illustrate the proposed method to estimate the average causal effect (ACE) with partially observed confounders in causal inference. Our framework applies to asymptotically linear ACE estimators, including the regression imputation, weighting, and matching estimators. We extend to the scenarios when both outcome and confounders are subject to missingness and when the data are missing not at random.

show abstract