| --- |
| license: cc-by-4.0 |
| task_categories: |
| - depth-estimation |
| - object-detection |
| - robotics |
| language: |
| - en |
| tags: |
| - mmWave-radar |
| - FMCW |
| - imaging |
| - depth-image |
| - semantics |
| size_categories: |
| - 10K<n<100K |
| --- |
| |
| # PanoRadar Dataset |
|
|
| PanoRadar is a multimodal indoor RF sensing dataset introduced in the MobiCom 2024 paper **“Enabling Visual Recognition at Radio Frequency.”** |
|
|
| It contains synchronized mmWave radar, LiDAR, and IMU measurements collected using a mobile robotic platform across **12 indoor buildings**. |
|
|
| The dataset supports research in RF imaging, depth estimation, semantic segmentation, object detection, robot localization, and multimodal perception. |
|
|
| ## Dataset Details |
|
|
| - **11,033 synchronized RF/LiDAR samples** |
| - **12 indoor buildings** |
| - **20,546 annotated object instances** |
| - Approximately **461 GB of processed data** |
| - **Curated by:** Haowen Lai, Gaoxiang Luo, Yifei Liu, and Mingmin Zhao, WAVES Lab, University of Pennsylvania |
| - **License:** CC BY 4.0 |
|
|
| ### Links |
|
|
| - **Repository:** https://github.com/penn-waves-lab/PanoRadar |
| - **Project:** https://penn-waves-lab.github.io/projects/panoradar |
| - **Paper:** *Enabling Visual Recognition at Radio Frequency*, ACM MobiCom 2024 |
| - **DOI:** `10.1145/3636534.3649369` |
| - **Demo:** https://youtu.be/dKyQ1XuPorU |
|
|
| ## Dataset Structure |
|
|
| The dataset contains two main components. |
|
|
| ### Raw Data |
|
|
| Raw recordings include: |
|
|
| - mmWave radar I/Q measurements |
| - synchronized LiDAR measurements |
| - synchronized IMU measurements |
| - timestamps |
| - moving and static robot trajectories |
|
|
| The raw data are intended for researchers working on radar signal processing, motion estimation, beamforming, and RF imaging. |
|
|
| ### Processed Data |
|
|
| Processed data include: |
|
|
| - 3D RF heatmaps |
| - LiDAR-derived range ground truth |
| - surface-normal ground truth |
| - glass masks |
| - semantic segmentation labels |
| - object-detection labels |
|
|
| These data can be used directly for training and evaluating machine-learning models without rerunning the full radar-processing pipeline. |
|
|
| ## Recommended Split |
|
|
| The original PanoRadar experiments use **leave-one-building-out (LOBO)** evaluation. |
|
|
| For each experiment: |
|
|
| - 11 buildings are used for training |
| - 1 building is held out for testing |
|
|
| This split is recommended when evaluating generalization to unseen environments. |
|
|
| ## Data Collection |
|
|
| Data were collected using a robotic platform equipped with: |
|
|
| - **TI AWR1843** mmWave FMCW radar |
| - **Ouster 64-beam LiDAR** |
| - IMU |
| - rotating radar assembly |
|
|
| The rotating radar forms a synthetic cylindrical array for panoramic 3D RF imaging. |
|
|
| LiDAR measurements provide geometric ground truth, while synchronized radar measurements are processed into 3D RF heatmaps. |
|
|
| ## Annotations |
|
|
| Range and geometric ground truth are derived primarily from synchronized LiDAR measurements. |
|
|
| Semantic segmentation and object-detection annotations were generated using a combination of machine-assisted labeling and human correction. |
|
|
| The dataset contains: |
|
|
| - **11,033 semantic and surface-normal annotations** |
| - **20,546 object instances** |
|
|
| ## Uses |
|
|
| PanoRadar is intended for research in areas including: |
|
|
| - RF and mmWave sensing |
| - 3D RF imaging |
| - depth and range estimation |
| - surface-normal estimation |
| - semantic segmentation |
| - object detection |
| - robot localization |
| - RF–LiDAR multimodal learning |
| - cross-building generalization |
|
|
| ## Limitations |
|
|
| The dataset was collected in **12 indoor buildings** and may not generalize directly to outdoor environments. |
|
|
| Results may also depend on the specific radar configuration, sensor mounting, and acquisition setup used in PanoRadar. |
|
|
| LiDAR is used as geometric ground truth but can be unreliable on transparent surfaces such as glass. Glass masks are included to help identify these regions. |
|
|
| Semantic and object annotations were produced using machine-assisted labeling with human correction and may contain annotation errors. |
|
|
| ## Citation |
|
|
| If you use PanoRadar, please cite: |
|
|
| ```bibtex |
| @inproceedings{PanoRadar, |
| author = {Lai, Haowen and Luo, Gaoxiang and Liu, Yifei and Zhao, Mingmin}, |
| title = {Enabling Visual Recognition at Radio Frequency}, |
| booktitle = {Proceedings of the 30th Annual International Conference on Mobile Computing and Networking}, |
| pages = {388--403}, |
| year = {2024}, |
| doi = {10.1145/3636534.3649369} |
| } |
| ``` |
|
|
| ## Contact |
|
|
| For questions or issues, please use the PanoRadar GitHub repository: |
|
|
| https://github.com/penn-waves-lab/PanoRadar |
