Midas: Generating mmWave Radar Data from Videos for Training Pervasive and Privacy-preserving Human Sensing Tasks
Use the links below to download the:
Folder description:
doppler_data/
used to store millimeter Doppler signals,mmw_preprocessing/
used to store millimeter wave signal processing files,models/
used to store weight files,video_data/
used to store video files.
File description:
train_encoder_decoder.py
used to train encoder-decoder,train_resnet34_classifier.py
used to train resnet34 classifier,train_vgg16_classifier.py
used to train vgg16 classifier.
Use the links below to download the:
- Real world Doppler dataset collected across 10 participants and 12 activities. Details found in paper here.
- Sample Videos for demo purposes.
Dataset directory structure:
video_data/
|----Participant_01/
| |----angle_0/
| |----action_01
| |----sample_video.mp4
| |----action_02
| |----sample_video.mp4
| |----...
| |----angle_45/
| |----...
| |----angle_90/
| |----...
|----...
radar_data/
|----Participant_01/
| |----angle_0/
| |----action_01
| |----doppler_gt.npy
| |----action_02
| |----doppler_gt.npy
| |----...
| |----angle_45/
| |----...
| |----angle_90/
| |----...
|----...
Data is divided into two parts, one is radar_data/
, which contains radar data. The other part is video_data/
, which contains video data, which we've only partially disclosed.
BibTex Reference:
...