Skip to content

Latest commit

 

History

History
40 lines (32 loc) · 1.93 KB

UseUWCR.md

File metadata and controls

40 lines (32 loc) · 1.93 KB
Raw

Use the Data from UWCR dataset for RAMP-CNN

This page introduces how to convert the data format of UWCR dataset to that required by RAMP-CNN model.

Download the UWCR dataset

Follow the instruction in the UWCR README and decompress the downloaded zip file.

Convert the Labels

To convert the annotations, you may use the function convert_annotations.py. An example of calling this function for the sequence '2019_04_09_cms1000' is presented in the script convertFormat.

    python convertFormat.py

The converted labels will be stored under the folder of "2019_04_09_cms1000" as "ramap_labels.csv".

Convert the Raw ADC Data

To convert the ADC data to the required format for training and testing (i.e.,, similar to the 'train_test_data.zip'), you may need to

  1. Convert the ADC data to the format of samples x antennas x chirps and save each frame to '.mat' file locally.

  2. Run the slice3d.py to generate RA slice, RV slice, VA slice. Note that please check if the input data format of slice3d.py is same to the ADC data in UWCR dataset

  3. Save ave the RA slice, RV slice, VA slice to '.npy' file in the folders RANPY, RVNPY, VANPY, respectively, following the below directory structure:

train_test_data
--date_1
----seq_1
--------camera image
--------RANPY
--------RVNPY
--------VANPY
--------ramp_labels.csv
----seq_2
    ...
--date_2
----seq_4
    ...

...
  1. You are done! Go ahead and use the new dataset.