ChangYuance / mmYodar

This is source code and dataset of the SECON 2023 paper "mmYodar: Lightweight and Robust Object Detection using mmWave Signals".

mmYodar

This is the source code and dataset of the SECON 2023 paper - "mmYodar: Lightweight and Robust Object Detection using mmWave Signals".

---

Overall, the mmYodar project includes four parts:
1.mmWave Signal Pre-processing Process
2.Real-time System
3.Camera and mmWave multimodal Dataset
4.Object Detection Network.

---

1.The mmWave signal pre-processing process of mmYodar is demonstrated in Radar_points_image folder.

Please run GetFiles.m to get radar points images which are the input of detection network. More details can be found in code.

Using your own mmWave binary files requires a deep understanding of mmWave data transmission and reception. Many parameters in the processing program need to be carefully modified accordingly.

Therefore, we provide pre-collected usable mmWave files. Baidu Netdisk:
https://pan.baidu.com/s/13XubeLqVNEAWVo74W8ct9A?pwd=ZXCV Extracted code：ZXCV

---

2.The real-time system of mmYodar is demonstrated in OnlineSystem folder.

Please run test.py for visual results. After initialization, you need to press the button DCA1000ARM and Trigger frame orderly.

Noted: Online system require installing mmWave studio software. Of course, you should have a mmwave radar.

---

3.The part of Camera and mmWave multimodal Dataset is demonstrated in Dataset folder. For privacy and security, please contact for a complete dataset [e-mail].

Therefore, we provide tested data.
https://pan.baidu.com/s/13iCNR4U6T8nXk-dXl_n5uA?pwd=l08i Extracted code：l08i

The dataset provides 16 scenario tests.

---

4.Object Detection Network.

Please run predict.py for visual results. You need to put yolo, utils, mini_net in the same directory.After that, download the trained weights.

Therefore, we provide trained weights.
https://www.aliyundrive.com/s/btVjpoXBVZv Extracted code：nr35

If you need to see the results of different test sets, please modify lines 26 and 27 in the predict.py, the data set is given in the third part.

---

Acknowledgements

The code for reading DCA1000 data in online system is partly borrowed from mmmesh.

The code is partly borrowed from https://github.com/bubbliiiing/yolo3-pytorch

---