mmRadPose

Welcome to the official repository for our mmRadPose datatset. A detailed explanation of the dataset is provided in the corresponding publication. If you use our dataset, please cite the following paper: "Advanced Millimeter Wave Radar-Based Human Pose Estimation Enabled by a Deep Learning Neural Network Trained With Optical Motion Capture Ground Truth Data" LUKAS ENGEL, JONAS MUELLER , EDUARDO JAVIER FERIA RENDON, EVA DORSCHKY, DANIEL KRAUSS, INGRID ULLMANN, BJOERN M. ESKOFIER, AND MARTIN VOSSIEK published in IEEE Journal of Microwaves (https://ieeexplore.ieee.org/document/10904474). L. Engel et al., "Advanced Millimeter Wave Radar-Based Human Pose Estimation Enabled by a Deep Learning Neural Network Trained With Optical Motion Capture Ground Truth Data," in IEEE Journal of Microwaves, doi: 10.1109/JMW.2025.3535525. keywords: {Radar;Pose estimation;Deep learning;Skeleton;Laser radar;Sensors;Radar measurements;Motion capture;Cameras;Synchronization;Human pose estimation;radar;optical motion capture;millimeter wave;target list;point cloud;machine learning;deep learning}, Datasets: POINTCLOUDS The pointcloud dataset contains radar pointclouds processed from raw radar data and skeleton ground truth data. mmRadPose_pointclouds/├── p1/ │ ├── angle0/ # angles │ │ ├── 0 # activities│ │ │ ├── 0 # recordings│ │ │ │ ├── skeleton.npy│ │ │ │ ├── targetlist_64.npy│ │ │ ├── 1 │ │ ├── 1│ │ ├── 2│ │ ├── ... │ │ ├── 10│ │ ├── 11│ ├── angle1/│ ├── angle2/├── p2 ├── ... ├── p12 RAW RADAR DATA The raw radar data dataset contains parsed () and cutted radar and skeleton ground truth (obained via OMC) data. radar folder: Contains the radar data in numpy compressed archive file format (e.g. data_cube_parsed_p1_an0_ac0_r0.npz) skeletons folder: Contains the skeleton GT data in numpy array file format (e.g. skeleton_p1_an0_ac0_r0.npy) File relation: All files have an unique ID (e.g. p1_an0_ac0_r0) to identify related radar and skeleton recordings. ID meaning: p X = participant X an X = angle X ac X = action X r X = recording X Edit: The skeleton files contain 26 keypoints provided by the OMC. However, only 24 keypoints are used in the paper.