Improving Depth Estimation Using Map-Based Depth Priors

Patil, Vaishakh; Liniger, Alexander; Dai, Dengxin; Gool, Luc Van

doi:10.1109/lra.2022.3146914

Cited by 4 publications

(1 citation statement)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Using prior maps to improve perception during deployment has been explored before, as demonstrated on depth estimation [38], semantic segmentation [39], 3D object detection [40]- [42], visual odometry [43], motion planning [44], [45], ego-lane detection [46] and change detection for road infrastructure either using 3D maps [47]- [49] or through image-to-image comparison [50], [51]. Despite the rapidly growing literature, existing approaches have limitations: they either rely on 3D maps (with significant variations in what map elements are used) [38], [40]- [42], [52]; only focus on updating static parts of the map [46]- [49]; require information from multiple prior traverses at test time [52]; or require pixel-level correspondences when using imageto-image setting [39], [43] In this work, for the task of dynamic vehicle detection in 2D, we consider a repeated route scenario for an autonomous vehicle. Thus, without needing a 3D map or pixel-level correspondences (especially under adverse conditions), we show how dynamic vehicle detections in the query image can be validated by comparing corresponding regions in the query and the reference image obtained through VPR.…”

Section: Prior Maps For Perceptionmentioning

confidence: 99%

Improving Worst Case Visual Localization Coverage via Place-Specific Sub-Selection in Multi-Camera Systems

Hausler

Garg

et al. 2022

IEEE Robot. Autom. Lett.

View full text Add to dashboard Cite

Can knowing where you are assist in perceiving objects in your surroundings, especially under adverse weather and lighting conditions? In this work we investigate whether a prior map can be leveraged to aid in the detection of dynamic objects in a scene without the need for a 3D map or pixel-level map-query correspondences. We contribute an algorithm which refines an initial set of candidate object detections and produces a refined subset of highly accurate detections using a prior map. We begin by using visual place recognition (VPR) to retrieve a reference map image for a given query image, then use a binary classification neural network that compares the query and mapping image regions to validate the query detection. Once our classification network is trained, on approximately 1000 query-map image pairs, it is able to improve the performance of vehicle detection when combined with an existing off-the-shelf vehicle detector. We demonstrate our approach using standard datasets across two cities (Oxford and Zurich) under different settings of train-test separation of map-query traverse pairs. We further emphasize the performance gains of our approach against alternative design choices and show that VPR suffices for the task, eliminating the need for precise ground truth localization.

show abstract