Millimeter-wave Object Recognition Dataset (MORD)

Citation Author(s):
Maloy Kumar
Mohammad Saeid
Submitted by:
Maloy Kumar Devnath
Last updated:
Thu, 05/11/2023 - 10:18
Data Format:
Research Article Link:
0 ratings - Please login to submit your rating.


In this investigation, the researchers have used a commercially available millimeter-wave (MMW) radar to collect data and assess the performance of deep learning algorithms in distinguishing different objects. The research looks at how varied ambiance factors, such as height, distance, and lighting, affect object recognition ability in both static and dynamic stages of the radar. The researchers examine five distinct objects comprised of four distinct materials—a dime, a quarter, a lead pencil, a plastic sheet, and a piece of wood—in order to capture data using the static radar mode under three different lighting conditions and two different heights. In order to capture data under three different lighting conditions and two distances, the researchers examine five distinct objects comprised of four materials in the moving radar mode: a UGV, a water bottle, a plastic sheet, paper, and clothing. The study's major goal is to identify potential obstacles in object recognition using MMW radar and investigate techniques to overcome them. The researchers hope to gain insight into how environmental factors affect object recognition performance. Overall, this study aims to improve knowledge of MMW radar-based object recognition and contribute to the creation of more robust and dependable deep-learning algorithms for this purpose.



Three discrete folders have been allocated for three distinct environments based on the criteria of distance or height. It holds true for static or dynamic modes of radar. The environment folders contain stored files. The environment folders are located within the distance and height folders. Each individual file corresponds to the specific data collected for the corresponding objects.


The dataset consists of four headers, each containing information about detected objects.

Each header contains the following information:

  • TimeStamp: the timestamp of the recognition in seconds and nanoseconds

  • frameNumber: the frame number of the recognition

  • NumberOfDetectedObject: the number of points detected in the frame from the object

For each detected object, the following information is provided:

  • x: the x-coordinate of the detected points from the object

  • y: the y-coordinate of the detected points from the object

  • z: the z-coordinate of the detected points from the object

  • r: the radial distance from the sensor to the detected points of the object

  • v: the velocity of the detected points from the object

  • snr: the signal-to-noise ratio of the detected point from the object

  • noise: the noise level of the detected points of the object

The x, y, z, r, v, snr, and noise values can be used to recognize the objects. The data file contains multiple records, each representing the detected points in a specific frame. The provided data can be used to train and test object recognition algorithms for MMW radar systems. The data can be used to develop and evaluate algorithms for detecting and tracking objects in different environments.


Funding Agency: 
This research is supported by the U.S. Army Grant #W911NF2120076, and NSF Research Experience for Under- graduates (REU) grant #CNS-2050999.



Submitted by Rahul Karanam on Mon, 05/15/2023 - 12:42

I do not understand. Will you please elaborate?

Submitted by Maloy Kumar Devnath on Fri, 05/19/2023 - 16:20