File size: 1,314 Bytes
42d074c
 
 
 
 
 
 
 
 
f31413e
 
42d074c
 
 
39a9baa
42d074c
 
 
63c8bab
f31413e
39a9baa
3a0bfe8
 
 
 
42d074c
f31413e
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
---
task_categories:
- feature-extraction
- sentence-similarity
tags:
- Online-HD-Map-Construction
- OpenStreetMap
- Computer-Vision
pretty_name: >-
  OSM Tag and Map Dataset used ín "SDTagNet: Leveraging Text-Annotated
  Navigation Maps for Online HD Map Construction" (NeurIPS 2025)
size_categories:
- 100M<n<1B
---
# Checkpoints, OSM Tags and Maps of SDTagNet: Leveraging Text-Annotated Navigation Maps for Online HD Map Construction (NeurIPS 2025)

<!-- Provide a quick summary of the dataset. -->

Checkpoints, OSM Tags and Maps of "SDTagNet: Leveraging Text-Annotated Navigation Maps for Online HD Map Construction" (NeurIPS 2025).
We publish the following data for easier reproducibility:
1. Trained checkpoints for SDTagNet in the near and far range setting
2. The pre-trained NLP encoder and a random initialized BERT encoder used for training from scratch
3. A pickle file of all unique tagsets in the OSM Planet map from 27.11.2024 (Warning: Uses 8-10x more RAM than its current size when unpickled)
4. The already processed contrastive pretraining dataset using the relevant tags for anchor-positive grouping, stored in sharded parquet format
5. The OSM SD maps for Argoverse 2 and nuScenes used during training

Project Page: [https://immel-f.github.io/SDTagNet/](https://immel-f.github.io/SDTagNet/)