ghua-ac / e-enf

Forked from the co-author's repo for CVPR 2023 paper on Event-ENF analysis

“Seeing” Electric Network Frequency from Events

Project Page | Paper | Data

“Seeing” Electric Network Frequency from Events

Lexuan Xu, Guang Hua, Haijian Zhang, Lei Yu, Ning Qiao

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR2023)

Prepare Data

Download the datasets from EV-ENFD.

1. Place one or several raw event files in the '.aedat4' format under the three scenarios in EV-ENFD into the 'Events/Raw/'.
2. Run 'Event_Process/aedat4_unpack_without_flir.py' to unpack '.aedat4' files in 'Raw', and the result will be saved in 'Events/Unpacked/dvSave-' (containing two folders: 'events' for unpacked events and 'event2ts' for frame-compressed event stream in time surface mode).
3. Replace 'Events/ENF_Reference' with the 'ENF_Reference' folder in EV-ENFD, where each '.wav' file contains grid voltage changes recorded by the transformer within an hour.

After performing the aforementioned operations, the contents of the 'Events' folder are as follows:

<project root>
  |-- Events
  |     |-- Raw
  |     |     |-- dvSave-2022_08_17_20_10_23.aedat4
  |     |     |-- dvSave-2022_08_17_20_24_41.aedat4
  |     |     |-- ...
  |     |-- Unpacked
  |     |     |-- dvSave-2022_08_17_20_10_23
  |     |     |    |-- events
  |     |     |    |-- event2ts
  |     |     |-- dvSave-2022_08_17_20_24_41
  |     |     |    |-- events
  |     |     |    |-- event2ts
  |     |     |-- ...     
  |     |-- ENF_Reference
  |     |     |-- 2022_08_17_Wed_17_00_00.wav
  |     |     |-- 2022_08_17_Wed_18_00_00.wav

Estimate Electric Network Frequency using E-ENF

To use the GUI interface shown above, run 'E_ENF/E_ENF(GUI)/ENF_match_GUI.py' and follow these steps:

1. Click on the 'Unpacked Events' button and select the desired event stream from the 'Events/Unpacked/dvSave-/events' folder (e.g., 'dvSave-2022_08_17_20_10_23/events') for extraction.
2. Select the real ground truth reference by clicking on the 'ENF_Reference Folder' button and choosing the 'Events/ENF_Reference' folder.
3. Press the 'Start' button to initiate the estimation of the ENF signal from the selected event stream. The estimated result will be displayed in the middle of the GUI.

Citation

Please cite our work if you use the code.

@inproceedings{xu2023seeing,
  title={"Seeing" Electric Network Frequency From Events},
  author={Xu, Lexuan and Hua, Guang and Zhang, Haijian and Yu, Lei and Qiao, Ning},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={18022--18031},
  year={2023}
}