Grilled Cheese

ExploreLog inSign up
Terms of UsePrivacy PolicyCommunity StandardsHelpGet the app

Grilled Cheese is a product of Village Compute

Version devBuilt at: 2026-10-10 01:38:52 EDT

Explore

PostsPeople
LatestRanked
@sciopentup.bsky.socialOct 7, 2026, 12:31 PM

A new V-SLAM method helps robots navigate shield tunnels using point and line features instead. Performance improves by up to 49%.
Full findings:
www.sciopen.com/article/10.2...

#SciOpen #TUP #VSLAM #RoboticsResearch @tsinghuauniversity.bsky.social

@wayne003.bsky.socialSep 9, 2026, 2:26 AM

Review of deep learning-enhanced visual SLAM 🤖 Learning boosts VSLAM robustness. ⚠️ Classical geometry fails in harsh scenes. 🔗 Five fusion interfaces categorized. 🧭 Future: foundation models, real-time deployment. DOI: https://doi.org/10.32604/cmc.2026.086341 #VSLAM #DeepLearning

Uncertainty-aware tracking and mapping workflow in UP-SLAM. The framework couples a tracking pipeline and a Gaussian mapping pipeline through a shared state of keyframes, poses, and features. Pose estimation and keyframe selection provide geometric constraints for online tracking, while probabilistic anchors and Gaussian map updates build a high-fidelity scene representation. Rendered image/depth predictions are assessed by feature and uncertainty consistency heads, allowing the system to suppress unreliable observations and improve robustness in trajectory estimation, map reconstruction, and novel-view rendering.
@wayne002.bsky.socialSep 9, 2026, 2:26 AM

Review of deep learning-enhanced visual SLAM 🤖 Learning boosts VSLAM robustness. ⚠️ Classical geometry fails in harsh scenes. 🔗 Five fusion interfaces categorized. 🧭 Future: foundation models, real-time deployment. DOI: https://doi.org/10.32604/cmc.2026.086341 #VSLAM #DeepLearning

Uncertainty-aware tracking and mapping workflow in UP-SLAM. The framework couples a tracking pipeline and a Gaussian mapping pipeline through a shared state of keyframes, poses, and features. Pose estimation and keyframe selection provide geometric constraints for online tracking, while probabilistic anchors and Gaussian map updates build a high-fidelity scene representation. Rendered image/depth predictions are assessed by feature and uncertainty consistency heads, allowing the system to suppress unreliable observations and improve robustness in trajectory estimation, map reconstruction, and novel-view rendering.
@w77576780.bsky.socialSep 9, 2026, 2:26 AM

Review of deep learning-enhanced visual SLAM 🤖 Learning boosts VSLAM robustness. ⚠️ Classical geometry fails in harsh scenes. 🔗 Five fusion interfaces categorized. 🧭 Future: foundation models, real-time deployment. DOI: https://doi.org/10.32604/cmc.2026.086341 #VSLAM #DeepLearning

Uncertainty-aware tracking and mapping workflow in UP-SLAM. The framework couples a tracking pipeline and a Gaussian mapping pipeline through a shared state of keyframes, poses, and features. Pose estimation and keyframe selection provide geometric constraints for online tracking, while probabilistic anchors and Gaussian map updates build a high-fidelity scene representation. Rendered image/depth predictions are assessed by feature and uncertainty consistency heads, allowing the system to suppress unreliable observations and improve robustness in trajectory estimation, map reconstruction, and novel-view rendering.