Text2Loc: 3D Point Cloud Localization from Natural Language
About
We tackle the problem of 3D point cloud localization based on a few natural linguistic descriptions and introduce a novel neural network, Text2Loc, that fully interprets the semantic relationship between points and text. Text2Loc follows a coarse-to-fine localization pipeline: text-submap global place recognition, followed by fine localization. In global place recognition, relational dynamics among each textual hint are captured in a hierarchical transformer with max-pooling (HTM), whereas a balance between positive and negative pairs is maintained using text-submap contrastive learning. Moreover, we propose a novel matching-free fine localization method to further refine the location predictions, which completely removes the need for complicated text-instance matching and is lighter, faster, and more accurate than previous methods. Extensive experiments show that Text2Loc improves the localization accuracy by up to $2\times$ over the state-of-the-art on the KITTI360Pose dataset. Our project page is publicly available at \url{https://yan-xia.github.io/projects/text2loc/}.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Global Place Recognition | KITTI360Pose (val) | Recall@10.32 | 15 | |
| Text-based position localization | KITTI360 Pose (test) | Localization Recall (k=1, ε < 5m)33 | 13 | |
| Localization | KITTI360Pose (val) | Recall @ 5m77 | 12 | |
| Localization | KITTI360Pose (test) | Recall @ 5m71 | 12 | |
| Text-to-point cloud localization | KITTI360 Pose (val) | Recall@k=1 (5m)37 | 11 | |
| Fine Localization | KITTI360Pose (val) | Recall@k=1 (5m Error)53 | 10 | |
| Fine Localization | KITTI360Pose (test) | Recall@1 (5m)0.47 | 10 | |
| Text-to-point-cloud-submap retrieval | KITTI360Pose (test) | Recall@10.28 | 8 | |
| Global Place Recognition | KITTI360 Pose (test) | Recall@128 | 5 |