15 skills found
NVlabs / GeomapnetGeometry-Aware Learning of Maps for Camera Localization (CVPR2018)
jike5 / P MapNetReceived by RAL
samarth-robo / MapNetCVPR 2018 Spotlight: MapNet: Geometry-Aware Learning of Maps for Camera Localization
vincentcartillier / Semantic MapNetNo description available
lehaifeng / MAPNetNo description available
ariseff / MapnetEstimate road layout attributes given street view imagery
jotaf98 / MapnetPyTorch implementation of the CVPR 2018 (oral) paper "MapNet: An Allocentric Spatial Memory for Mapping Environments" (Henriques and Vedaldi)
jpdong-xjtu / DAMap[ICCV2025] DAMap: Distance-aware MapNet for High Quality HD Map Construction
shuowang666 / SQD MapNet[ECCV2024] SQD-MapNet: Stream Query Denoising for Vectorized HD-Map Construction
m5823779 / Pose EstimationA Convolutional Neural Network for Real Time robot pose estimation by RGB Image
srama2512 / Mapnet PytorchUnofficial PyTorch implementation of MapNet: An Allocentric Spatial Memory for Mapping Environments
sarankhaliq2326 / NUST CLCWe investigate why ORB-SLAM is missing fre- quently occurring loop closures. Investigating failures in loop closure module of ORB-SLAM is challenging since it consists of multiple building blocks. Our meticulous investigations have revealed few interesting findings. Contrary to reported results, ORB-SLAM frequently misses large fraction of loop closures on public (KITTI, TUM RGB-D) datasets. One common assumption is, in such scenarios, the visual place recognition (vPR) block of the loop closure module is unable to find a suitable match due to extreme conditions (dynamic scene, viewpoint/scale changes). We report that native vPR of ORB SLAM is not the sole reason for these failures. Interestingly, recent deep vPR alternatives achieve impressive matching performance. Replacing native vPR with these deep alternatives will partially improve the loop closure performance of visual SLAM. The other part of the problem is the subsequent relative pose estimation module between the matching pair. Surpris- ingly, using off-the-shelf SIFT based relative pose estimation manages to close most of the loops missed by ORB-SLAM. This finding advocates the use of multiple features for different SLAM modules instead of relying on a one-feature-for-all strategy. Furthermore, the performance of deep relocalization methods (MapNet) is worse than classic methods even in case of loop closures scenarios. This finding further supports the fundamental limitation of deep relocalization methods recently diagnosed. Finally, we provide a challenging loop closure dataset which enabled these findings. The dataset can be used to test future loop closing routines against challenging yet commonly occurring indoor navigation scenarios.
GMU-vision-robotics / Mapnet Habitat NavigationCode for paper 'Simultaneous Mapping and Target Driven Navigation' on Habitat environment.
MapNetNZ / Mapnetnz.github.ioMAPNet website
bryantky / MAPNetNo description available
