The raw dataset is simulated by 24Ghz FMCW radar, containing 11 daily human activities. Standing in a fixed position while rotating his body (B); kicking (K), punching (P), grabbing an object (G), walking back and forth in front of the radar (W), standing up from chair (SU), sitting down on chair (SD), stands up from chair to walk (STW), walks to sit on chair (WTS), walks to fall on the ground (WTF), standing up from ground to walk (FTW).

The RF dataset can be used to develop and test detection algorithms for the 3.5 GHz CBRS or similar bands where the primary users of the band are federal incumbent radar systems. The dataset consists of synthetically generated radar waveforms with added white Gaussian noise. The RF dataset is suitable for development and testing of machine/deep learning detection algorithms. A large number of parameters of the waveforms are randomized across the dataset. Due to its large size, the dataset is divided into groups, and each group consists of multiple files. For more information about the dataset, refer to: R. Caromi, M. Souryal, and T. Hall, "RF Dataset of Incumbent Radar Systems in the 3.5 GHz CBRS Band," Journal of Research of the National Institute of Standards and Technology. (in press). In addition, the metadata of the dataset is summarized in "Data Dictionary of 3.5 GHz Radar Waveforms" [pdf] accompanying the data. For more information about the motivation behind this RF dataset, refer to: T. Hall, R. Caromi, M. Souryal, and A. Wunderlich, "Reference Datasets for Training and Evaluating RF Signal Detection and Classification Models," to appear in Proc. IEEE GLOBECOM Workshop on Advancements in Spectrum Sharing, Dec. 2019.

This dataset consists of raw ADC readings from a 3 transmitter 4 receiver 77GHz FMCW radar, together with synchronized RGB camera and depth (active stereo) measurements. The data is grouped into 4 distinct radar configurations: - "indoor" configuration with range <14m - "30m" with range <38m - "50m" with range <63m - "high_res" with doppler resolution of 0.043m/s # Related code https://github.com/moodoki/radical_sdk # Hardware Project Page https://publish.illinois.edu/radicaldata

The data is collected using millimeter-wave radar IWR1642 in receiver-only configuration (passive mode) for four cases namely, single AWR1642 as an ECM jammer, and single AWR2944 as an ECM jammer, two AWR1642 as jammers, and three AWR devices as jammers at shorter distances. The data is useful for estimating the angle of arrival in single and multi-jammer scenarios.

This dataset contains the weather radar (WXR) data collected during the High Altitude Ice Crystals - High Ice Water Content (HAIC-HIWC) project that took place in Cayenne, French Guiana. Some of the data files provide 4D tracks of the aircraft that can be played within Google Earth. The other data files contain the radar data. The data files are arranged by flight and are in zip, kml, and kmz format.

This dataset contains the weather radar (WXR) data collected during the High Altitude Ice Crystals - High Ice Water Content (HAIC-HIWC) project that took place in Darwin, Australia. Some of the data files provide 4D tracks of the aircraft that can be played within Google Earth. The other data files contain the radar data. The data files are arranged by flight and are in zip, kml, and kmz format.

The RadarScenes data set (“data set”) contains recordings from four automotive radar sensors, which were mounted on one measurement-vehicle. Images from one front-facing documentary camera are added.

The data set has a length of over 4h and in addition to the point cloud data from the radar sensors, semantic annotations on a point-wise level from 12 different classes are provided.

In addition to point-wise class labels, a track-id is attached to each individual detection of a dynamic object, so that individual objects can be tracked over time.

Structure of the Data Set

The data set consist of 158 individual sequences. For each sequence, the recorded data from radar and odometry sensors are stored in one hdf5 file. Each of these files is accompanied by a json file called “scenes.json” in which meta-information are stored. In a subfolder, the camera images are stored as jpg files.

Two additional json files give further meta-information: in the "sensor.json" file, the sensor mounting position and rotation angles are defined. In the file "sequences.json", all recorded sequences are listed with additional information, e.g. about the recording duration.

sensors.json

This file describes the position and orientation of the four radar sensors. Each sensor is attributed with an integer id. The mounting position is given relative to the center of the rear axle of the vehicle. This allows for an easier calculation of the ego-motion at the position of the sensors. Only the x and y position is given, since no elevation information is provided by the sensors. Similarly, only the yaw-angle for the rotation is needed.

sequences.json

This file contains one entry for each recorded sequence. Each entry is built from the following information: the category (training or validation of machine learning algorithms), the number of individual scenes within the sequence, the duration in seconds and the names of the sensors which performed measurements within this sequence.

scenes.json

In this file, meta-information for a specific sequence and the scenes within this sequence are stored.

The name of the sequence is listed within the top-level dictionary, the group of this sequence (training or validation) as well as the timestamps of the first and last time a radar sensor performed a measurement in this sequence.

A scene is defined as one measurement of one of the four radar sensors. For each scene, the sensor id of the respective radar sensor is listed. Each scene has one unique timestamp, namely the time at which the radar sensor performed the measurement. Four timestamps of different radar measurement are given for each scene: the next and previous timestamp of a measurement of the same sensor and the next and previous timestamp of a measurement of any radar sensor. This allows to quickly iterate over measurements from all sensors or over all measurements of a single sensor. For the association with the odometry information, the timestamp of the closest odometry measurement and additionally the index in the odometry table in the hdf5 file where this measurement can be found are given. Furthermore, the filename of the camera image whose timestamp is closest to the radar measurement is given. Finally, the start and end indices of this scene’s radar detections in the hdf5 data set “radar_data” is given. The first index corresponds to the row in the hdf5 data set in which the first detection of this scene can be found. The second index corresponds to the row in the hdf5 data set in which the next scene starts. That is, the detection in this row is the first one that does not belong to the scene anymore. This convention allows to use the common python indexing into lists and arrays, where the second index is exclusive: arr[start:end].

radar_data.h5

In this file, both the radar and the odometry data are stored. Two data sets exists within this file: “odometry” and “radar_data”.

The “odometry” data has six columns: timestamp, x_seq, y_seq, yaw_seq, vx, yaw_rate. Each row corresponds to one measurement of the driving state. The columns x_seq, y_seq and yaw_seq describe the position and orientation of the ego-vehicle relative to some global origin. Hence, the pose in a global (sequence) coordinate system is defined. The column “vx” contains the velocity of the ego-vehicle in x-direction and the yaw_rate column contains the current yaw rate of the car.

The hdf5 data set “radar_data” is composed of the individual detections. Each row in the data set corresponds to one detection. A detection is defined by the following signals, each being listed in one column:

• timestamp: in micro seconds relative to some arbitrary origin
• sensor_id: integer value, id of the sensor that recorded the detection
• range_sc: in meters, radial distance to the detection, sensor coordinate system
• azimuth_sc: in radians, azimuth angle to the detection, sensor coordinate system
• rcs: in dBsm, RCS value of the detection
• vr: in m/s. Radial velocity measured for this detection
• vr_compensated in m/s: Radial velocity for this detection but compensated for the ego-motion
• x_cc and y_cc: in m, position of the detection in the car-coordinate system (origin is at the center of the rear-axle)
• x_seq and y_seq in m, position of the detection in the global sequence-coordinate system (origin is at arbitrary start point)
• uuid: unique identifier for the detection. Can be used for association with predicted labels and debugging
• track_id: id of the dynamic object this detection belongs to. Empty, if it does not belong to any.
• label_id: semantic class id of the object to which this detection belongs. passenger cars (0), large vehicles (like agricultural or construction vehicles) (1), trucks (2), busses (3), trains (4), bicycles (5), motorized two-wheeler (6), pedestrians (7), groups of pedestrian (8), animals (9), all other dynamic objects encountered while driving (10), and the static environment (11)

Camera Images

The images of the documentary camera are located in the subfolder “camera” of each sequence. The filename of each image corresponds to the timestamp at which the image was recorded.

The data set is a radar data set. Camera images are only included so that users of the data set get a better understanding of the recorded scenes. However, due to GDPR requirements, personal information was removed from these images via re-painting of regions proposed by a semantic instance segmentation network and manual correction. The networks were optimized for high recall values so that false-negatives were suppressed at the cost of having false positive markings. As the camera images are only meant to be used as guidance to the recorded radar scenes, this shortcoming has no negative effect on the actual data.

Tools

Some helper tools - including a viewer - can be found in the python package radar_scenes. Details can be found here: https://github.com/oleschum/radar_scenes

Publications

Previous publications related to classification algorithms on radar data already used this data set:

• Scene Understanding With Automotive Radar; https://ieeexplore.ieee.org/document/8911477

• Semantic Segmentation on Radar Point Clouds, https://ieeexplore.ieee.org/document/8455344

• Off-the-shelf sensor vs. experimental radar - How much resolution is necessary in automotive radar classification?, https://ieeexplore.ieee.org/document/9190338

• Detection and Tracking on Automotive Radar Data with Deep Learning, https://ieeexplore.ieee.org/document/9190261

• Comparison of random forest and long short-term memory network performances in classification tasks using radar, https://ieeexplore.ieee.org/document/8126350

License

The data set is licensed under Creative Commons Attribution Non Commercial Share Alike 4.0 International (https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode). Hence, the data set must not be used for any commercial use cases.

Disclaimer

That the data set comes "AS IS", without express or implied warranty and/or any liability exceeding mandatory statutory obligations. This especially applies to any obligations of care or indemnification in connection with the data set. The annotations were created for our research purposes only and no quality assessment was done for the usage in products of any kind. We can therefore not guarantee for the correctness, completeness or reliability of the provided data set.

This dataset consists of Level II weather radar data collected from Next-Generation Radar (NEXRAD) stations located in the contiguous United States, Alaska, Hawaii, U.S. territories and at military base sites. NEXRAD is a network of 160 high-resolution Doppler weather radars operated by the NOAA National Weather Service (NWS), the Federal Aviation Administration (FAA), and the U.S. Air Force (USAF). Doppler radars detect atmospheric precipitation and winds, which allow scientists to track and anticipate weather events, such as rain, ice pellets, snow, hail, and tornadoes, as well as some non-weather objects like birds and insects. NEXRAD stations use the Weather Surveillance Radar - 1988, Doppler (WSR-88D) system. This is a 10 cm wavelength (S-Band) radar that operates at a frequency between 2,700 and 3,000 MHz. The radar system operates in two basic modes: a slow-scanning Clear Air Mode (Mode B) for analyzing air movements when there is little or no precipitation activity in the area, and a Precipitation Mode (Mode A) with a faster scan for tracking active weather. The two modes employ nine Volume Coverage Patterns (VCPs) to adequately sample the atmosphere based on weather conditions. A VCP is a series of 360 degree sweeps of the antenna at pre-determined elevation angles and pulse repetition frequencies completed in a specified period of time. The radar scan times 4.5, 5, 6 or 10 minutes depending on the selected VCP. The NEXRAD products are divided into multiple data processing levels. The lower Level II data contain the three meteorological base data quantities at original resolution: reflectivity, mean radial velocity, and spectrum width. With the advent of dual polarization beginning in 2011, additional base products of differential reflectivity, correlation coefficient and differential phase are available. Level II data are recorded at all NWS and most USAF and FAA WSR-88D sites. From the Level II quantities, computer processing generates numerous meteorological analysis Level 3 products. NEXRAD data are acquired by the NOAA National Centers for Environmental Information (NCEI) for archiving and dissemination to users. Data coverage varies by station and ranges from June 1991 to 1 day from present. Most stations began observing in the mid-1990s, and most period of records are continuous.

This dataset contains the weather radar (WXR) data collected during the High Ice Water Content (HIWC) Radar Study project, onboard the NASA DC-8 aircraft, that took place in Fort Lauderdale, Florida; Palmdale, California; and Kona, Hawaii. The radar measurements contained in this dataset have been processed to remove the instrument properties from the airborne radar data and thereby produce scientific measures of the atmosphere and the HIWC conditions. There are 3 different file types (RadProd, bitmap, and KMZ) containing radar measurements with each file type containing all of the flights in one zip file.

Each .zip archive contains a collection of .dat files obtained via a Ground Penetrating Radar (GPR).The numerical values in each file are obtained by defining sampling points with a grid of equally spaced lines (50 mm distance between each other) on each coordinate axis.Archive names follow the convention "GPR_ DATE.zip" where DATE can be either 30-08-2017 or 31-08-2017 while .mat file names follow the convention "GPR_X#_Y#.dat" where X# spans from X0 to X13 and Y# from Y0 to Y22.

The up-to-date weather radar from the FMI radar network is available as Open Data. The data contain both single radar data along with composites over Finland in GeoTIFF and HDF5-formats. Available composite parameters consist of radar reflectivity (DBZ), rainfall intensity (RR), and precipitation accumulation of 1, 12, and 24 hours. Single radar parameters consist of radar reflectivity (DBZ), radial velocity (VRAD), rain classification (HCLASS), and Cloud top height (ETOP 20). Raw volume data from singe radars are also provided in HDF5 format with ODIM 2.3 conventions. Radar data becomes available as soon as it's received from the radar and pre-processed into deliverable formats. Typically the most recent radar data was collected less than 5 minutes ago.

RadDet is a challenging public dataset for radar spectrum detection. It comprises a large corpus of radar signals occupying a wideband spectrum across diverse radar density environments and signal-to-noise ratio settings. This dataset was published together with our paper at the 2025 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2025) in Hyderabad, India.

Further information about the RadDet dataset is available in the paper:

Z. Huang, S. Denman, A. Pemasiri, T. Martin and C. Fookes, RadDet: A Wideband Dataset for Real-Time Radar Spectrum Detection, ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Hyderabad, India, 2025, pp. 1-5, doi: 10.1109/ICASSP49660.2025.10887772.

This dataset consists of Level III weather radar products collected from Next-Generation Radar (NEXRAD) stations located in the contiguous United States, Alaska, Hawaii, U.S. territories and at military base sites. NEXRAD is a network of 160 high-resolution Doppler weather radars operated by the NOAA National Weather Service (NWS), the Federal Aviation Administration (FAA), and the U.S. Air Force (USAF). Doppler radars detect atmospheric precipitation and winds, which allow scientists to track and anticipate weather events, such as rain, ice pellets, snow, hail, and tornadoes, as well as some non-weather objects like birds and insects. NEXRAD stations use the Weather Surveillance Radar - 1988, Doppler (WSR-88D) system. This is a 10 cm wavelength (S-Band) radar that operates at a frequency between 2,700 and 3,000 MHz. The radar system operates in two basic modes: a slow-scanning Clear Air Mode (Mode B) for analyzing air movements when there is little or no precipitation activity in the area, and a Precipitation Mode (Mode A) with a faster scan for tracking active weather. The two modes employ nine Volume Coverage Patterns (VCPs) to adequately sample the atmosphere based on weather conditions. A VCP is a series of 360 degree sweeps of the antenna at pre-determined elevation angles and pulse repetition frequencies completed in a specified period of time. The radar scan times 4.5, 5, 6 or 10 minutes depending on the selected VCP. During 2008, the WSR-88D radars were upgraded to produce increased spatial resolution data, called Super Resolution. The earlier Legacy Resolution data provides radar reflectivity at 1.0 degree azimuthal by 1 km range gate resolution to a range of 460 km, and Doppler velocity and spectrum width at 1.0 degree azimuthal by 250 m range gate resolution to a range of 230 km. The upgraded Super Resolution data provides radar reflectivity at 0.5 degree azimuthal by 250 m range gate resolution to a range of 460 km, and Doppler velocity and spectrum width at 0.5 degree azimuthal by 250 m range gate resolution to a range of 300 km. Super resolution makes a compromise of slightly decreased noise reduction for a large gain in resolution. In 2010, the deployment of the Dual Polarization (Dual Pol) capability to NEXRAD sites began with the first operational Dual Pol radar in May 2011. Dual Pol radar capability adds vertical polarization to the previous horizontal radar waves, in order to more accurately discern the return signal. This allows the radar to better distinguish between types of precipitation (e.g., rain, hail and snow), improves rainfall estimates, improves data retrieval in mountainous terrain, and aids in removal of non-weather artifacts. The NEXRAD products are divided in two data processing levels. The lower Level II data are base products at original resolution. Level II data are recorded at all NWS and most USAF and FAA WSR-88D sites. From the Level II quantities, computer processing generates numerous meteorological analysis Level III products. The Level III data consists of reduced resolution, low-bandwidth, base products as well as many derived, post-processed products. Level III products are recorded at most U.S. sites, though non-US sites do not have Level III products. There are over 40 Level III products available from the NCDC. General products for Level III include the base and composite reflectivity, storm relative velocity, vertical integrated liquid, echo tops and VAD wind profile. Precipitation products for Level III include estimated ground accumulated rainfall amounts for one and three hour periods, storm totals, and digital arrays. Estimates are based on reflectivity to rainfall rate (Z-R) relationships. Overlay products for Level III are alphanumeric data that give detailed information on certain parameters for an identified storm cell. These include storm structure, hail index, mesocyclone identification, tornadic vortex signature, and storm tracking information. Radar messages for Level III are sent by the radar site to users in order to know more about the radar status and special product data. NEXRAD data are provided to the NOAA National Climatic Data Center for archiving and dissemination to users. Data coverage varies by station and ranges from May 1992 to 1 day from present. Most stations began observing in the mid-1990s, and most period of records are continuous.

RadChar is a synthetic radar signal dataset designed to facilitate the development of multi-task learning models. Unlike existing datasets that only provide labels for classification tasks, RadChar provides labels that support both classification and regression tasks in radar signal recognition. This makes it the first multi-task labelled dataset of its kind released to help the research community to advance machine learning for radar signal characterisation.

Further information about the RadChar database is available in our paper:

Z. Huang, A. Pemasiri, S. Denman, C. Fookes and T. Martin, Multi-Task Learning For Radar Signal Characterisation, 2023 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW), Rhodes Island, Greece, 2023, pp. 1-5, doi: 10.1109/ICASSPW59220.2023.10193318

Ground Penetrating Radar

## Overview

Ground Penetrating Radar is a dataset for object detection tasks - it contains Hyperbola annotations for 520 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

Radar sensors have a long tradition in advanced driver assistance systems (ADAS) and also play a major role in current concepts for autonomous vehicles. Their importance is reasoned by their high robustness against meteorological effects, such as rain, snow, or fog, and the radar’s ability to measure relative radial velocity differences via the Doppler effect. The cause for these advantages, namely the large wavelength, is also one of the drawbacks of radar sensors. Compared to camera or lidar sensor, a lot more surfaces in a typical traffic scenario appear flat relative to the radar’s emitted signal. This results in multi-path reflections or so called ghost detections in the radar signal. Ghost objects pose a major source for potential false positive detections in a vehicle’s perception pipeline. Therefore, it is important to be able to segregate multi-path reflections from direct ones.

Here we present a dataset with detailed manual annotations for different kinds of ghost detections. We hope that our dataset encourages more researchers to engage in the fields of multi-path object suppression or exploitation.

Paper: https://ieeexplore.ieee.org/document/9636338 (10.1109/IROS51168.2021.9636338)

Accompanying github repository: https://github.com/flkraus/ghosts

Documentation and further information can be found on the github repo.

This software tool generates simulated radar signals and creates RF datasets. The datasets can be used to develop and test detection algorithms by utilizing machine learning/deep learning techniques for the 3.5 GHz Citizens Broadband Radio Service (CBRS) or similar bands. In these bands, the primary users of the band are federal incumbent radar systems. The software tool generates radar waveforms and randomizes the radar waveform parameters. The pulse modulation types for the radar signals and their parameters are selected based on NTIA testing procedures for ESC certification, available at http://www.its.bldrdoc.gov/publications/3184.aspx. Furthermore, the tool mixes the waveforms with interference and packages them into one RF dataset file. The tool utilizes a graphical user interface (GUI) to simplify the selection of parameters and the mixing process. A reference RF dataset was generated using this software. The RF dataset is published at https://doi.org/10.18434/M32116.

The main radar covering the Toronto Area is the King City Doppler Dual-Polarization C-Band radar (43.96388, -79.57416). Other nearby radars include the Exeter Doppler C-Band radar (43.37027,-81.38416) and the Buffalo Doppler Dual-Polarization S-Band radar (42.94889,-78.73667) from the United States. Though the primary radar for the project is the King City radar, the raw data from all three radars are included in their native format (IRIS or Nexrad Level 2) and are intended for radar specialists. The data is available from May 1 2015 to Sept 30 2015. The scan strategy for each radar is different, with at least 10 minute scan cycles or better. The user should consult with a radar specialist for more details. Reflectivity and radial velocity images (presented as a pair) for the lowest elevation angle (0.5o) centred on a 128 km x 128 km box around from the King City radar are provided for general use. Besides their normal use as precipitation observations, they are particularly useful to identify Lake Breezes as weak linear reflectivity and as radial velocity discontinuity features for the entire period. The target providing the radar returns are insects. Analysis indicates the presence of Lake Breezes on 118 days, only 35 days did not have any kind of lake breeze-like features. Daily movies have been created. The format of the single images is PNG and the movies is an animated GIF. The data is organized by radar and by day in the following structure. The raw data is organized in the following directory structure: RADAR -> -> YYYY -> YYYYMM -> YYYYMMDD; where YYYY is the appropriate year (always 2015), MM is the month, DD is the day. The images are organized as: RADAR -> QUICKLOOK -> YYYY -> YYYYMM -> YYYYMMDD; where YYYY is the year, MM is the month, DD is the day. The movies are located in the YYYYMMDD folders and are named as YYYYMMDD_movie.gif

The ER-2 X-band Radar (EXRAD) IMPACTS dataset consists of radar reflectivity and Doppler velocity estimates collected by the EXRAD onboard the NASA ER-2 high-altitude research aircraft. These data were gathered during the Investigation of Microphysics and Precipitation for Atlantic Coast-Threatening Snowstorms (IMPACTS) field campaign. IMPACTS was a three-year sequence of winter season deployments conducted to study snowstorms over the U.S. Atlantic Coast (2020-2023). The campaign aimed to (1) Provide observations critical to understanding the mechanisms of snowband formation, organization, and evolution; (2) Examine how the microphysical characteristics and likely growth mechanisms of snow particles vary across snowbands; and (3) Improve snowfall remote sensing interpretation and modeling to significantly advance prediction capabilities. The EXRAD IMPACTS dataset files are available from January 25, 2020, through March 2, 2023, in HDF-5 format.

Radar

## Overview

Radar is a dataset for object detection tasks - it contains Radar annotations for 1,510 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).